2024 Experimental Psychology PDF

EXPERIMENTAL PSYCHOLOGY 01. Experimental Psychology and the Scientific 4.) Overconfidence Bias - feels more Method confident about conclusions than is...

EXPERIMENTAL PSYCHOLOGY 01. Experimental Psychology and the Scientific 4.) Overconfidence Bias - feels more Method confident about conclusions than is warranted by available data. This form ✣ Experimental Psychology is the study of of nonscientific inference can result in psychological issues that uses experimental erroneous conclusions when failed to procedures. It is is an area of psychology that recognize the limitations of supporting utilizes scientific methods to research the mind data. and behavior. ✣ Science connotes “content” and “process”. ✣ Alfred North Whitehead’s - scientific ✣ Psychology is the science of behavior, thus it mentality assumes that behavior follows a relies on scientific methodology and techniques natural order and can be predicted. This in gathering data and analyzing behavior. assumption is essential to science. There is no ✣ Methodology consists of the scientific point to using the scientific method to gather and techniques we use to collect and evaluate data. analyze data if there is no implicit order. ✣ Data are the facts we gather using scientific methods. ✣ Empirical - when data are observed or experienced. We get knowledge from: – Galileo’s empirical approach was 1.) Philosophy 2.) Folk wisdom 3.) Concepts superior to Aristotle’s common sense method. 4.) Law and theories 5.) Common sense – Galileo correctly concluded that light objects fall as rapidly as heavy ones in a ✣ Heider called non scientific data gathering vacuum. commonsense psychology. – This approach uses non scientific ✣ Law - consists of statements generally sources of data and non scientific inference. An expressed as equations with few variables that everyday example is believing that “opposites have overwhelming empirical support. attract.” ✣ Theory - an interim explanation; a set of related statements used explain and predict ✣ Nonscientific inference is the nonscientific phenomena. Theories integrate diverse data, use of information to explain or predict explain behavior, and predict new instances of behavior. behavior. ✣ Good thinking - critical to the scientific 1.) Gambler’s Fallacy - people misuse data method. We engage in good thinking when data to estimate the probability of an event, collection and interpretation are systematic, like when a slot machine will pay off. objective, and rational. 2.) Overuse Trait Explanations - explain ✣ Principle of Parsimony - we prefer the others' behavior. Often make simplest useful explanation. For example, unwarranted dispositional attributions Crandall (1988) showed that a social contagion and underuse situational information. model of bulimia was more parsimonious than This bias can reduce the accuracy of competing explanations. explanations and predictions. 3.) Stereotyping - falsely assume that ✣ Sir Karl Popper - proposed that science specific behaviors cluster together. advances by revising theories based on the Stereotypes ignore individual “weight of evidence”. Science is self-correcting differences. as scientific explanations and theories are hypotheses and establish challenged, and revised or replaced. cause-and-effec trelationships. ✣ Principle of Modus Tollens - allows us to Experimentation is not always possible disprove statements using a single, contrary because our predictions must be observation. We can never prove a statement testable. because a contradictory observation might be found later. ✣ Replication - an exact or systematic repetition of a study. It increases our confidence in experimental results by adding to the weight of supporting evidence. Four Main Objectives of Science Requirements for an experiment 1.) Description - a systematic and unbiased ✣ We must be able to manipulate the account of observed characteristics of independent variable and measure its effect on behaviors. the dependent variable. Ethical concerns or 2.) Prediction - the capability of knowing technological limitations may prevent in advance when certain behaviors experimentation. should occur. ✣ An experiment requires that we create at least 3.) Explanation - knowledge of the two treatment conditions and randomly assign conditions that reliably produce a subjects to these conditions. In psychology behavior. experiments, we control extraneous variables so 4.) Control - the use of scientific we that we can measure “what we intend to knowledge to influence behavior. measure.” ✣ An experiment attempts to establish a cause- ✣ Applied research - addresses real-world and-effect relationship between the antecedent problems like how to improve student conditions (IV) and subject behavior (DV). graduation rates. ✣ Experiments establish a temporal ✣ Basic research - tests theories and explains relationship, because causes must precede psychological phenomena like helping behavior. effects. However, not all prior events are causes. Main Tools of Psychological Science ✣ Pseudoscience - any field of study that gives 1.) Observation - the systematic noting and the appearance of being scientific, but has no recording of events. Observations must true scientific basis and has not been confirmed be objective so that there can be strong using the scientific method. Modern agreement among raters. pseudosciences include past life regression, – Systematic means that the procedures reparenting, and rebirthing. are consistently applied. The events or their signs must be observable. 02. Research Ethics 2.) Measurement - assigns numbers to objects, events, or their characteristics. ✣ Research Ethics - a framework of values This is an inherent feature of within which we conduct research. Ethics help quantitative research. Baron and researchers identify actions we consider good colleagues (1985) measured anger and and bad, and explain the principles by which we depression using numerical scales. make responsible decisions in actual situations. 3.) Experimentation - the process we use ✣ Institutional review boards (IRBs) - to test the predictions we call composed of laypeople and researchers who evaluate research proposals to make sure that Who is targeted by APA ethical guidelines? they follow ethical standards. APA ethical guidelines apply to psychologists – IRBs protect the safety of research and students when they assume the role of participants. Their first task is to decide whether psychologists during research or practice. a proposed study increases participants’ risk of injury since psychological research can cause ✣ Deception - may be used when it is the best physical and/or psychological discomfort. way to obtain information. It may not be used to – As researchers, we must accurately estimate minimize the participants’ perception of risk or the degree of risk in our research. We typically exaggerate their perception of potential benefits. do this by reading the literature and consulting Which steps must researchers take if deception with colleagues. is used? Subjects must be allowed to withdraw – IRBs will also help researchers estimate the from the experiment at any time and should degree of risk involved in their studies. never face coercion to remain. – Studies that place subjects at risk increase the chance of harm compared with not participating ✣ Debriefing - the experimenter should provide in the study. Minimal risk studies do not increase full disclosure after either their personal the likelihood of injury. participation or the completion of the entire – IRBs should approve an “at risk” study when a study. risk/benefit analysis determines that risks to – Debriefing involves explaining the true nature participants are outweighed by gains in and purpose of the experiment. knowledge. – Debriefing is an essential component of good experimental research. We must offer our Proposed Three Principles of Belmont Report participants a full explanation of our study any (1979) time that we use deception. 1.) Respect for persons: individuals have the right of self-determination (basis of ✣ Principle of Full Disclosure - explaining the informed consent). true nature and purpose of the study to the 2.) Beneficence: minimize harm and subject at the end of their participation or at the maximize potential benefits (basis of completion of the entire experiment. risk/benefit analysis). – In debriefing, an experimenter discloses the 3.) Justice: fairness in both the burdens and true nature and purpose of the study to the benefits of research. subject and solicits subjects’ questions at the end of the experiment. What is informed consent? A subject or guardian agrees in writing to the subject’s ✣ Confederate - an experimenter’s accomplice. participation after relevant details of the Use of a confederate is deceptive because experiment have been explained. This subjects are led to believe that the confederate is description may include risks and benefits, but another subject, experimenter, or bystander, does not extend to deception or the hypothesis. when he or she is actually part of the – Perhaps the most important principle built into experimental manipulation. ethics codes is the right of a participant to refuse to be in the study or discontinue participation. ✣ Anonymity - subjects are not identified by – Ethical researchers, therefore, cannot coerce name. participants to agree to be in the study or prevent – Researchers achieve anonymity by collecting participants from discontinuing the study. data without names and assigning code numbers. ✣ Confidentiality - data are securely stored be given to those who made a major contribution and only used for the purpose explained to the to the research or writing. Researchers should subject. not take credit for the same research more than – They achieve confidentiality by storing data in once. a locked safe and only using the data for the – The ethical solution is to cite original purposes explained to the participants. publications when republishing data in a journal article or republishing journal articles in an How do psychologists protect the welfare of edited volume. animal subjects? 03. Alternatives to Experimentation: ✣ Animal Welfare - the humane care and Nonexperimental Designs treatment of animals. Institutions that conduct animal research must establish an Institutional ✣ Nonexperimental Approaches - do not Animal Care and Use Committee (IACUC) to create levels of an independent variable nor evaluate animal research before it is conducted. randomly assign subjects to these levels. – The IACUC must determine that the – They are used where experiments are not researchers have explored all alternatives and ethicalor possible, or where we want to test have documented that there are no other feasible hypotheses in realistic conditions. alternatives. ✣ Animal rights - the position that sensate ✣ Internal Validity - the degree to which a species (those that can feel pain and suffer) have researcher can establish a causal relationship equal value and rights to humans. between the independent and dependent variables. ✣ Scientific Fraud - involves falsifying or – An experiment has high internal validity when fabricating data. A researcher’s graduation, we can demonstrate that only the antecedent tenure, promotion, funding, or reputation may conditions are responsible for group differences motivate researchers to commit fraud. in behavior. – An internally valid experiment allows us to Main lines of defense against fraud: draw cause-and-effect conclusions. 1.) Peer Review - process filters submitted manuscripts so that only 15-20% of Why do experiments often achieve higher articles are printed. internal validity than nonexperimental studies? 2.) Replication - where researchers attempt Laboratory experiments are often higher in to reproduce the findings of others, is internal validity because of their control of the second line of defense. extraneous variables. Researchers create levels 3.) Competition by colleagues for scarce of the IV and use procedures like matching and resources, while a cause of fraud, is the random assignment to conditions. third line of defense. ✣ External Validity - the degree to which ✣ Plagiarism - misrepresenting someone’s research findings can be generalized to other “ideas, words, or written work” as your own. It settings and individuals. is a form of fraud, in which an individual claims false credit for another’s ideas, words, or written Why might nonexperimental studies achieve work. higher external validity than laboratory experiments? Nonexperimental studies are more Which ethical issues may be involved in frequently conducted in real-world settings with research reports? Authorship credit should only a more diverse sample of participants than 3.) What’s It Like To Be Diagnosed With A experiments. Terminal Disease When A Person Becomes A Parent? The Conflict Of Birth And Death Can’t ✣ Degree of manipulation Be Generalized, But Research Can Record of antecedent conditions - Emotions And Experiences. concerns assignment of subjects to antecedent ✣ Case Studies - a researcher compiles a conditions created for the descriptive study of a subject's experiences, experiment. observable behaviors, and archival records kept by an outside observer. – It is an in-depth examinations of people, groups of people or institutions rather than a sweeping statistical survey. – It is a method used to narrow down a very broad field of research into one easily researchable topic. Strengths 1.) Source of inferences, hypotheses, and ✣ Degree of Imposition of Units - how much theories. Source of therapy techniques you limit a subject’s responses on the DV. 2.) Allow study of rare phenomena – It refers to the extent to which experimental 3.) Provide exceptions to accepted ideas, manipulations or interventions are imposed on theories, and practices participants. This concept is closely related to 4.) Persuasive and motivational value the idea of experimental control and the balance (advertising) between internal validity and external validity. Limitations Examples of Nonexperimental Approach 1.) Representativeness of sample ✣ Phenomenology involves a subject's 2.) Completeness of data description of personal subjective experience. 3.) Reliance on retrospective data – Phenomenological studies examine human experience through the descriptions that are ✣ Deviant Case Analysis - researchers provided by the people involved. examine differences between deviant and normal – It is a study that attempts to understand individuals to identify etiological factors. people's perceptions, perspectives and – This approach may also be applied to understandings of a particular situation (or nonclinical issues such as social trends and adult phenomenon). morale. Examples: Retrospective Data - recollections of past 1.) The Experiences Of Every War Survivor Or events that are collected in the present. While War Veteran Are Unique. Research Can your childhood memories constitute Illuminate Their Mental States And Survival retrospective data, your undergraduate portfolio Strategies In A New World. does not since it was collected in the past. 2.) Losing Family Members To Covid-19 Hasn’t Been Easy. A Detailed Study Of Survivors And Risk People Who’ve Lost Loved Ones Can Help This information may be compromised by faulty Understand Coping Mechanisms And memory, current mood, and the retrieval cues Long-Term Traumas. that are present when you are asked to recall an ✣ Archival Study - a descriptive method where event. researchers reexamine data that were collected for other purposes. ✣ Field Studies - nonexperimental studies – For example, universities collect a wealth of conducted in the field (real-life settings). The data through surveys like the Graduating Senior experimenter does not manipulate antecedent Questionnaire (GSQ) and interviews. conditions. Field studies range from low-low to low-high. ✣ Qualitative Research - obtains data consisting of words instead of numbers. This Naturalistic Observation - examines information is obtained through self-reports, subjects’ spontaneous behavior in their actual personal narratives, and expression of ideas, environments and may obtain more memories, feelings, and thoughts. representative behavior than experiments. This method can achieve high levels of external Why is the rise of qualitative research validity. important? The increased use of qualitative research may represent a paradigm shift—a Reactivity - subjects alter their behavior when change in attitudes, values, beliefs, methods, and they know that they are being observed. For procedures accepted during a specific time example, your baby sister stops saying period. “Gramma” when you place the phone near her mouth. Qualitative research is invaluable in studying contextual phenomena, behavior that can only be ✣ Participant-Observer Study - involves field understood within its context. For example, we observation in which the researcher is part of the might examine the meaning of religious faith for studied group. This approach contrasts with patients facing impending surgery. naturalistic observation, where the researcher does not interact with research subjects to avoid ✣ Empirical Phenomenology - might rely on reactivity. an experimenter’s private experiences or other experiential data: Which ethical problems complicate 1.) the researcher’s self-reflection on relevant participant-observer studies? The main experiences problems are invasion of privacy, not telling 2.) participants’ oral or written descriptions of people that you are studying their behavior, and their experiences pretending to be a group member. Pretending to 3.) accounts from literature, poetry, visual art, be a group member (e.g., a researcher pretending television, theatre, and previous to be a weightlifter) is a serious problem that phenomenological (and other) research requires careful planning. 04. Alternatives to Experimentation: Surveys ✣ Field experiments - experiments conducted and Interviews in real-life settings. – Field studies are nonexperimental designs ✣ Survey Research - obtains data about used in real-life settings and include naturalistic opinions, attitudes, preferences, and behaviors observation, unobtrusive measures, using questionnaires or interviews. participant-observer studies, and surveys. – The survey approach allows researchers to study private experience, which cannot be directly observed. Advantages 2. Avoid double-barreled (compound) questions We can efficiently collect large amounts of that require responses about two or more data. unrelated ideas. Anonymous surveys can increase the accuracy 3. Use exhaustive response choices. of answers to sensitive questions. Surveys can allow us to draw inferences about ✣ Nominal Scale - assigns items to two or the causes of behavior and can complement more distinct categories that can be named using laboratory and field experiments. a shared feature, but does not measure their magnitude. For example, you can sort professors Limitation into exciting and dull categories. The survey approach does not allow us to test ✣ Ordinal Scale - measures the magnitude of hypotheses about causal relationships because the dependent variable using ranks, but does not we do not manipulate independent variables and assign precise values. For example, marathon control extraneous variables. contestants may finish from first place to last place. What are the major steps in constructing ✣ Interval Scale - measures the magnitude of surveys? the DV using equal intervals between values 1.) Identify specific research objectives. with no absolute zero point. For example, 2.) Decide on the degree of imposition of units Fahrenheit or Centigrade temperatures, and (degree of response restriction). Sarnoff and Zimbardo’s (1961) 0-100 scale. 3.) Decide how you will analyze the survey data. ✣ Ratio Scale - measures the magnitude of the dependent variable using equal intervals Major Question Types: between values and an absolute zero. This scale allows us to state that a 2-meter board is twice as – Closed questions (structured questions) - long as a 1-meter board. For example, distance can be answered using a limited number of in meters. alternatives and have a high imposition of units. For example, “How many songs did your How to select measurement scales? roommate illegally download this month?” – The best type of scale depends on the variable – Open-ended questions (open questions) - you are studying and the level of precision you require that participants respond with more than desire. a yes or 1-10 rating and have a low imposition – Since psychological variables like traits, of units. For example, “Why did your choose attitudes, and preferences represent a continuous your major?” dimension, several levels of measurement “fit” equally well. How do researchers analyze data from each – When working with variables like sociability, question type? The number or percent of psychologists often select the highest scale since responses can be reported for closed questions. it provides more information and allows analysis – Open-ended questions can be analyzed using using more powerful statistics. content analysis, like Yepez’s INTERSECT, in which responses are assigned to categories using What should be considered when creating objective rules. survey items? – Subjects decide to refuse to answer surveys Three concerns when constructing during the start or first few questions. questions. – Engage subjects from the start by asking 1. Keep items simple and unambiguous, and interesting questions they will not mind avoid double negatives. answering. – The first survey question should be: ✣ Structured Interviews - questions are asked 1. Relevant to the survey’s central topic the same way each time. This provides more 2. Easy to answer usable, quantifiable data. 3. Interesting ✣ Unstructured Interviews - the interviewer 4. Answerable by most respondents can explore interesting topics as they arise. 5. Closed format These data may not be usable for content – Whenever possible, use commonly used analysis. response options. – Avoid value-laden questions that might ✣ Population - consists of all people, animals, make a response seem embarrassing. or objects that share at least one characteristic. ✣ Sample - a subset of the population of ✣ Response Styles - tendencies to respond to interest (the population we are studying). questions or test items without regard to their actual wording. What are two advantages of probability – People differ in their willingness to answer, sampling over nonprobability sampling? position preference, and yea-saying and 1.) A probability sample is more likely to nay-saying. represent the population (external validity) than a nonprobability sample. 1.) Willingness to answer - the tendency to 2.) We know the exact odds of members of the guess or omit items when unsure. population being included in our sample. This 2.) Position preference - selecting an tells us whom the sample represents. answer based on its position. For example, students choosing “c" on Main Probability Sampling Methods multiple-choice exams. – Probability Sampling involves choosing 3.) Manifest content - the plain meaning of subjects using a procedure that allows us to the words printed on the page. While we calculate their chance of selection from a expect subjects to respond to the population. Researchers use four forms of manifest content of questionnaires, they probability sampling: simple random sampling, may ignore it when answering questions systematic random sampling, stratified random about their feelings or attitudes. sampling, and cluster sampling. 4.) Yea-saying - agreeing with an item regardless of its manifest content. Four Main Probability Sampling Methods 5.) Nay-saying - disagreeing with an item 1.) Simple Random Sampling - the most basic regardless of its manifest content. form of probability sampling. Researchers select 6.) Context effects - changes in question subjects using an unbiased procedure (like a interpretation due to their position table of random numbers) in which each subject within a survey. This problem is has an equal chance of being included in our especially likely when two questions are sample. related and not separated by buffer items 2.) Systematic Random Sampling - all (unrelated questions). members of a population are known, a 7.) Social desirability response set - researcher can randomly select every nth person representing ourselves in a socially from the population, where n is determined by appropriate fashion when responding to the size of the population and desired sample a question’s latent content (underlying size. meaning). For example, you may dress 3.) Stratified Random Sampling - researchers formally for a job interview instead of identify and randomly sample from each wearing your favorite jeans. important population subgroup (like “Democrats” and “Republicans”) in the same ✣ Correlational Studies - the researcher proportion as found in the population. This examines the strength of relationships between method increases the precision of our estimates variables by strengthening how changes in one and is used extensively by professional polling variable are associated with changes in another operations. variable. 4.) Cluster Sampling - researchers randomly A correlation indicates the extent to which one sample from groups (like zip codes and variable (x) is related to another variable (y). counties) that already exist. This allows us to The magnitude and the direction of the sample efficiently from a limited number of relationship between two variables are indicated locations. by a correlation coefficient. Correlation coefficient may be positive (+) or – Nonprobability Sampling chooses subjects negative (-) and range from -1.00 (perfect using a biased procedure. Quota sampling, negative correlation) to 1.00 (perfect positive convenience sampling, purposive sampling, and correlation) if the correlation coefficient has no snowball sampling are four forms of sign in front of it (.80), a positive relationship is nonprobability sampling. In each case, indicted. researchers do not select subjects at random. A negative correlation coefficient will be Since they do not know a person’s odds of being preceded by a negative sign(-.80). A correlation selected, they do not know whom the sample coefficient of 0.00 indicates no relationship represents. between variables. Four Main Nonprobability Sampling Methods Properties of Correlation 1.) Quota Sampling - researchers recruit 1.) Linearity - means how the relationship subjects until they meet a predetermined quota between x and y can be plotted as a line that mirrors the makeup of the population. (linear relationship) or a curve 2.) Convenience Sampling - researchers recruit (curvilinear relationship). subjects who are convenient to study (such as 2.) Sign - refers to whether the correlation students who showed up for class). This is a coefficient is positive or negative. weak form of sampling since researchers do not 3.) Magnitude - the strength of the control a sample’s representativeness. correlation coefficient, ranging from -1 3.) Purposive Sampling - researchers select to +1. nonrandom samples based on the specific 4.) Probability - the likelihood of obtaining purpose of the study (studying the success of a a correlation coefficient of this new employee training program in a company’s magnitude due to chance. sales and human resources departments). 5.) Scatterplots - a graphic display of pairs 4.) Snowball Sampling - researchers locate of data points on the x and y axes. A individuals who meet the sample criteria (animal scatterplot illustrates the linearity, sign, rights activists) and ask them to find or lead magnitude, and probability (indirectly) them to additional individuals. This method is of a correlation. mainly used when sampling very small, 6.) Coefficient of Determination (r2) - uncommon, or unique populations when estimates the amount of variability that researchers do not know who the population can be explained by a predictor variable. members are or how to contact them. For example, Chaplin et al. (2000) showed that handshake firmness 05. Alternatives to Experimentation: accounted for 31% of the variability of Correlational and Quasi-Experimental Designs first impression positivity.  7.) Researchers use multiple correlation Advantages (R) when they want to know whether Allows applied research when experiments not there is a relationship among three or possible. more variables. We could measure age, Threats to internal validity can (sometimes) be television watching, and vocabulary and assessed. find that R = +.61. Practical and more feasible than true 8.) We should compute a partial experiments, especially in clinical settings. correlation when we want to hold one Some generalizability variable (age) constant to measure its influence on a correlation between two Disadvantages other variables (television watching and Difficult to make clear cause-and-effect vocabulary). statements. 9.) Researchers use multiple regression to Statistical analysis can be difficult. predict behavior measured by one Most statistical analyses assume variable based on scores on two or more randomness. other variables. We could estimate Can not randomize assignment to groups. vocabulary size using age and television watching as predictor variables. ✣ Developmental Designs - used to study development or changes in behavior. Describe ✣ Comparative Studies - this examines the relationship between age and other variables differences between intact groups on some dependent variable of interest. The difference Three main types between these studies and experimental studies 1.) Cross-sectional - uses a separate group of lies in the researchers ability to manipulate the participants for each age group being compared independent variable. Different groups measured once and compared – There is no manipulation of the independent to each other variable. Frequently, the independent variable is Between subjects design some inherent characteristics of the subjects, Most commonly used such as personality type, educational level, or Short period of time medical conditions. No real commitment Gather all data at one time ✣ Methodological Studies - these are 2.) Longitudinal - same participants are concerned with the development, testing, and observed over time. evaluation of research instruments and methods. Assesses stability of traits This is justified that the instruments are should Individuals are compared to self throughout be valid and reliable measures of the variables of repeated measures over time interest. Within subjects design 3.) Cohort-sequential - measure groups of ✣ Quasi-Experimental Design - almost “true” participants as they age. experiments but lack of control over assignment Combines the best features of both of participants. Independent variable cannot be longitudinal and cross-sectional designs manipulated (inherent confound); Studies specific age groups over time – Subject variable Both between and within subjects design – Time could be variable (Developmental) – Random variable already present How do quasi-experiments differ from actual experiments? Quasi means “seeming like.”  Quasi-experiments superficially resemble 6. Determining the basic research approach experiments, but lack their required 7. Identifying the population and sample manipulation of antecedent conditions and/or 8. Designing the data collection plan random assignment to conditions. 9. Selecting or developing data collection They may study the effects of preexisting instruments antecedent conditions—life events or subject 10. Choosing the method of data analysis characteristics—on behavior. 11. Implementing the research plan 12. Interpreting the results ✣ Quasi-experimental Designs Ex Post Facto Design - a quasi-experimental ✣ Steps in Experimental Research study examining how an independent variable, 1. State the research problem present prior to the study in the participants, 2. Determine if experimental methods apply affects a dependent variable. 3. Specify the independent variable(s) – A quasi-experimental study simply 4. Specify the dependent variable(s) means participants are not randomly assigned. 5. State the tentative hypotheses Nonequivalent Control Group Design - a 6. Determine measures to be used treatment group and a comparison group are 7. Pause to consider potential success compared using pretest and posttest measures. 8. Identify intervening (extraneous) variables – If the two groups are similar in their 9. Formal statement of research hypotheses pretest scores prior to treatment but differ in 10. Design the experiment their posttest scores following treatment, 11. Final estimate of potential success researchers can more confidently make a claim 12.Conduct the study as planned about the effect of treatment. 13.Analyze the collected data Pretest-Posttest Design - usually a 14. Prepare a research report quasi-experiment where participants are studied before and after the experimental manipulation. ✣ The Research Question – It explores the effects of an event The foundation of the research process (treatment) by comparing behavior before and It all begins with a question. after the event (treatment). For example: Source: Practice GRE test 1 🡪 six-week preparation – Curiosity – Information Gaps – Controversy course 🡪 Practice GRE test 2. – Replication – Literature Review – Other People 06. The Research Problem ✣ Research Topic - the broad general area ✣ Stages of the Scientific Method: steps within expected to investigate. It is a broad idea or the research process. concept from which many problems may be Question Identified ⇢ Hypothesis Formed ⇢ delineated. Research Plan ⇢ Data Collected ⇢ Results Analyzed ⇢ Conclusions ✣ Research Problem - a situation or circumstance that requires a solution to be ✣ Steps of Research described, explained, or predicted. It is an 1. Identify the research question unsatisfactory situation that wants you to 2. Initial review of literature confront. 3. Distilling the question to a researchable problem ✣ If there is a knowledge gap in an area that 4. Continued review of literature need to be investigated, the research problem 5. Formulation of hypothesis identifies this gap. Where as the research topic is simply a broad area of interest, the research – It translates the research purpose into a clear problem identifies what is problematic about prediction of the expected results or outcome of that topic. the study. – A hypothesis is an explanation of a ✣ Research Statement - a statement specifies relationship between two or more variables. exactly what is being studied. – It is the research’s prediction of the outcome of the research study. Six Elements of Research Statement – It translates the research purpose into a clear 1.) Information about the research topic that prediction of the expected results or outcome of provoked the study. the study. 2.) The scope of the problem (e.g.. how many people are affected by it). ✣ Experimental Hypothesis - a tentative 3.) Why it is important to study the explanation of an event or a behavior. It is a problem. statement that predicts the effect of an 4.) How Psychology would be influenced independent variable on a dependent variable. by the study. – For example, cognitive behavior therapy 5.) General characteristics of the population (CBT) produces less relapse than of interest. antidepressants. 6.) The overall goal or aim of the study or the question to be answered. ✣ Non-experimental Hypothesis - predicts how variables (events, traits, or behaviors) might Criteria for Selecting a Problem be correlated, but not causally related. Interest – For example, red-haired patients receive less – Most important relief from pain medication than blonde patients. Significance – Theoretical value Importance of Research Hypothesis – Practical value It provides direction for the type of research – Timeliness (i.e. design, sampling, data collection, etc.). – External review Suggests the type of statistical analysis to be Manageability used in the study. – Expertise, time, resources Identifies the variables to be manipulated – Free from personal bias and/or measured. ✣ Problem Distillation - the process of How to state a Research Hypothesis refining the question or idea into a problem and Research hypothesis should be stated clearly, making it sufficiently specific so that it is concisely, measurably, and in the present tense. amenable to investigation. This process should For a hypothesis to be stated clearly, concisely, lead to the development of a “statement of the and measurably, these criteria should be problem” that is clear, concise, and definitive. considered: 1.) A relationship should be addressed in 07. The Basics of Experimentation each hypothesis. 2.) A hypothesis must be capable of being ✣ Research Hypothesis - a hypothesis is an true or false, which is a property of explanation of a relationship between two or synthetic statements. more variables. 3.) The variable/condition/relationship must – It is the research’s prediction of the outcome of be testable or measurable. the research study.  4.) It is preferred to have a simple Operational Definitions hypothesis over one requiring many ✣ Experimental Group - the subjects in an supporting assumptions. experiment who are exposed to the treatment (independent variable). Types of Research Hypothesis – Also called the experimental condition. – The group being studied and compared to the ✣ Simple - it predicts the relationship between control group. one independent variable and one dependent ✣ Control Group - are not exposed to the variable. independent variable. Example: Newborns of smoking mothers (I.V) – Results are compared to those of the have lower birth weight (D.V.) than those of experimental group. non-smoking mothers. – Also called the control condition. Example: Lower levels of exercise postpartum (I.V) will be associated with greater weight Experimental Variables retention (D.V). ✣ Independent variable (IV) – the controlled factor in an experiment ✣ Complex - it predicts the relationship – hypothesized to cause an effect on another between 2 or more I.V. and 2 or more D.V. Variable Example: Structured preoperative support (I.V) – causes something to happen is more effective in reducing surgical patient’s – the variable manipulated by the experimenter perception of pain (D.V) and request of – the variable which should change the analgesics (D.V) than structured post operative dependent variable support (I.V). – variable is controlled by the experimenter ✣ Associative - it predicts an association ✣ Dependent variable (DV) between the I.V. and the D.V. without specifying – the measured facts either a directional or causal relation. – hypothesized to be affected Example: Maternal age (I.V.) is associated with – the experimental variable which is affected by pregnancy complications (D.V). the independent variable – the “effect variable” ✣ Causal - it predicts a cause-and effect – the outcome of the experiment relationship between the I.V. and D.V. – the variable being observed and measured Example: Older mothers (I.V.) give birth to newborns with lower age (D.V.) than those of Knowing the Difference younger mother (I.V). Find DV first by asking: – “What is the researcher ✣ Null - it predicts no relationship between I.V. measuring or looking for in and D.V. It is used when statistical testing this study?” procedures are applied to the data. Next, find IV by asking: Example: There is no relationship between – “What do the researchers maternal smoking and newborn’s birth weight. hope will cause the DV in this study?” ✣ Alternative - it is the opposite of the null Verify with an If/Then Statement: hypothesis. So, it predicts a relationship between – If this (IV) THEN this happens (DV). the I.V. and D.V. – If my subject drinks an energy drink (IV) Example: All the previous examples are THEN they should get a surge in energy (DV) alternative hypotheses. OR – Both are important in a study but they are frequently at odds with one another in planning They are testing the effect of (IV) on (DV). and designing a study. Good Way to Remember: An IV in your arm – Internal validity is considered the basic causes something to happen (DV). minimum for experimental research. ✣ Confounding Variables - variables, other Internal Validity than the independent variable, which could — This is the basic minimum without which any inadvertently influence the dependent variable study is not interpretable – “Outside factors” that could have caused your — Particularly important in experimental studies results. — Did, in fact, the experimental treatment (X) – Need to be controlled/eliminated in order to produce a change in the dependent variable (Y)? draw a true, cause-effect relationship in the — To gain internal validity, the researcher experiment. attempts to control everything and eliminate – Many confounding variables can be eliminated possible extraneous influences. through random assignment. — Lends itself to highly controlled, laboratory settings. Environmental Differences - any differences in the experiment’s Threats to Internal Validity conditions (between the experimental History – events occurring during the and control groups). experiment that are not part of the treatment – Differences include temperature, Maturation – biological or psychological lighting, noise levels, distractions, etc. processes within participants that may change – Ideally, there should be a minimum of due to the passing of time, e.g., aging, fatigue, environmental differences between the hunger. two groups. Testing – the effects of one test upon Expectation Effects (Participant/ subsequent administrations of the same test. Researcher Bias) - any changes in an Instrumentation – changes in testing experiment’s results due to the subject instruments, raters, or interviewers including or researcher anticipating certain lack of agreement within and between observers. outcomes to the experiment Statistical Regression – the fact that groups – Change in DV produced by subject’s selected on the basis of extreme scores are not as expectancy that change should happen extreme on subsequent testing. – Researcher favoring one group over Selection bias – identification of comparison another. groups in other than a random manner. Experimental mortality – loss of participants Research Validity from comparison groups due to nonrandom ✣ Internal Validity – the validity of findings reasons. with the research study; the technical soundness Interaction among factors – factors can of a study, particularly concerned with the operate together to influence experimental control of extraneous influences that might results. affect the outcome ✣ External Validity – the degree to which the External Validity findings can be inferred to the population of — Generalizability of results... to what interest or to other populations or settings; the populations, settings, or treatment variables can generalizability of the results. the results be generalized? — Concerned with real-world applications — What relevance do the findings have beyond – However, predictive validity does not define a the confines of the experiment? measure or construct. — External validity is generally controlled by selecting subjects, treatments, experimental Construct Validity - the degree to which the situations, and tests to be representative of some independent and dependent variables accurate larger population. reflect or measure what they are intended to — Random selection is the key to controlling measure. most threats to external validity. – Extraneous Variables: confounding variables that may be a source of invalidity can threaten Types of External Validity construct validity. Population Validity - refers to the extent to – Reactivity and Random Error which the results can be generalized from the experimental sample to a defined population ✣ Reliability - the consistency of behavioral Ecological Validity - refers to the extent to measures which the results of an experiment can be generalized from the set of environmental Types of Reliability: conditions in the experiment to other Test-Retest: giving the same test twice in environmental conditions. succession over a short time interval in order to measure consistency (using a correlation Threats to External Validity coefficient to measure consistency). Interaction effects of testing – the fact that Parallel Forms: giving two versions of a test the pretest may make the participants more on two testing occasions to determine whether aware of or sensitive to the upcoming treatment. they result in consistent scores. Selection bias – when participants are selected Split-Half: dividing test items from a single in a manner so they are not representative of any test into two arbitrary groups and correlating the particular population. resulting scores after administration—if the Reactive effects of experimental setting – the correlation is sufficiently high, then test fact that treatments in constrained laboratory reliability is confirmed (this also establishes the settings may not be effective in less constrained, equivalency of your test items). real-world settings. Multiple-treatment interference – when Validity – Does the experiment measure and participants receive more than one treatment, the predict what it is supposed to? effects of previous treatments may influence Reliable – If repeated, will we get similar subsequent ones. results? Types of Validity Methods of Control Predictive validity - checking the truth of an ✣ Physical Manipulation - best way to control observation by comparing it to another criterion extraneous variables. that is thought to measure the same thing – Researcher attempts to control all aspects of – In predictive validity, the relation between two the research, except the experimental treatment. scores is typically assessed by a statistic termed – Difficult to control all variables. the correlation coefficient (e.g., Pearson’s product-moment correlation coefficient) ✣ Selective Manipulation - intent is to – The better the prediction of the observation, increase likelihood that treatment groups are the greater the predictive validity of the similar at the beginning of study. predictor score. – Matched pairs and block designs – Counterbalanced designs ✣ Statistical Techniques - applied when physical manipulation or selective manipulation is not possible. – Differences among treatment groups are known to exist at the beginning of study. (Groups may differ on initial ability) – Analysis of covariance (ANCOVA). (Adjusts scores at the end of the study based upon initial differences.) 08. Basic Experimental Design Common Sources of Error ✣ Principles Of Experimental Design - Many possible sources of error can cause the experimental design controls background results of a research study to be incorrectly variability so that systematic effects of interpreted. The following sources of error are treatments can be observed. more specific threats to the validity of a study than those described previously. Three basic principles Selected examples: 1.) Control by Matching - known sources of Hawthorne Effect - a specific type of reactive variation may be eliminated by matching. effect in which merely being a research – Eliminating genetic variation participant in an investigation may affect – Compare animals from the same family of behavior. It suggests that, as much as possible, mice. participants should be unaware they are in – Eliminating district or school effects an experiment and unaware of the hypothesized – Compare students within districts or schools outcome. – However matching is limited Placebo Effect - participants may believe that – Matching is only possible on observable the experimental treatment is supposed to characteristics change them, so they respond to the treatment – Perfect matching is not always possible with a change in performance. – Matching inherently limits generalizability by John Henry Effect - a threat to internal validity removing (possibly desired) variation. wherein research participants in the control – Matching ensures that groups compared are group try harder just because they are in the alike on specific known and observable control group. characteristics (in principle, everything we have Rating Effect - variety of errors associated thought of). with ratings of a participant or group. – Wouldn’t it be great if there were a method of – Halo effect – Overrater error – Underrater making groups alike on not only everything we error – Central tendency error have thought of, but everything we didn’t think Experimenter Bias Effect - The intentional or of too? unintentional influence that an experimenter – There is such a method (researcher) may exert on a study. Replication - repeating the experiment to 2.) Control by Randomization determine if similar results are found. If so, the – Matching controls for the effects of variation research is considered reliable. due to specific observable characteristics. – Example: Does Vitamin C really prevent – Randomization controls for the effects all colds? (observable or unobservable, known or unknown) characteristics. – Randomization makes groups equivalent (on – Same or different subjects in different average) on all variables (known and unknown, treatment conditions. observable or not). Determined by – Randomization also gives us a way to assess – The nature of the hypothesis whether differences after treatment are larger – Information the researcher is seeking than would be expected due to chance. – Practical problems in running the experiment – Random assignment is not assignment with no Questions to consider particular rule. It is a purposeful process. – How many independent variables are there? – Assignment is made at random. This – How many treatment conditions are needed? does not mean that the experimenter writes – Will we use same or different subjects? down the names of the varieties in any order that occurs to him, but that he carries out a physical [ BETWEEN-SUBJECTS DESIGN ] experimental process of randomization, using Different subjects take part in each means which shall ensure that each variety will condition of the experiment have an equal chance of being tested on any Comparison between the behaviors of particular plot of ground (Fisher, 1935, p. 51) different groups of subjects – Random assignment of schools or classrooms Two Groups - Two Independent Groups is not assignment with no particular rule. It is a – Experimental group-Control group purposeful process. – 2 Experimental group – Assignment of schools to treatments is Two Matched Groups made at random. This does not mean that the – Experimental group-Control group experimenter assigns schools to treatments in – 2 Experimental group any order that occurs to her, but that she carries Multiple Groups out a physical experimental process of randomization, using means which shall ensure SELECTING AND RECRUITING SUBJECTS that each treatment will have an equal chance of The more the sample resembles the whole being tested in any particular school (Hedges, population, the more likely is it that the 2007). behaviors of the sample mirrors that of the population. 3.) Control by Statistical Adjustment How many subjects are needed for – Control by statistical adjustment is a form of between-subjects design? pseudo-matching. – at least 15 to 20 subject in each treatment – It uses statistical relations to simulate condition matching. – More comfortably, 30 per treatment – Statistical control is important for increasing precision but should not be relied upon to [ BETWEEN-SUBJECTS DESIGNS control biases that may exist prior to assignment. (1 IV, 2 GROUPS) ] – Statistical control is the weakest of the three How were the subject assigned? experimental design principles because its – Two Independent Groups validity depends on knowing a statistical model – Matched Groups for responses. What are the levels of IV? – Experimental group – Control group Design 08. 2. Basic Between-Subjects Design – Two Experimental Group Design EXPERIMENTAL DESIGN TWO INDEPENDENT GROUPS General structure of the experiment Grouped using random assignment – Number of treatment conditions  What are the possible treatment conditions Matching is possible (level of the IV) for 2 independent groups? Experimental group-Control group; ASSIGNING SUBJECTS Experimental condition-Control condition; Random number table – just pointing numbers Treatment-No treatment; however, “no in a pool of numbers in an unbiased way treatment” does not really mean “no treatment” Use of block randomization – use of treatment all the time. blocks. Two Experimental group (No Control group) – gathers more precise information; CHOOSING TREATMENTS different values of the IV. How does one choose how many treatments to When can 2 independent groups be used? use? – If there is only one independent variable – Theoretical basis – If randomization can be assumed successful – Equal or proportional intervals What will I gain by adding these extra TWO MATCHED GROUPS conditions to the experiment? Does/must/can not use random assignment As a general rule, select the simplest design 2 groups of subjects assigned by the researcher that will make an adequate test of your by equating them on a particular characteristic hypothesis. that might affect the DV. Matching can be done before or after the PRACTICAL LIMITS experiment. More subject are needed How to match? Takes more time – Precision Matching: matched pairs have More complicated statistical procedures identical scores Use of Pilot study (mini experiment) pretest – Range Matching: matched pair fall in a selected levels of an IV before conducting the specified range. Note: Some subjects have to be actual experiment. discarded – Rank-order Matching: subject scores are 09. Between-Subjects Factorial Design ranked; adjacent scores are paired. When can 2 matched groups be used? More Than One Independent Variable – Presence of a strong extraneous variable – In real life, variable rarely occur alone. – Need for more efficient measures of the IV’s – Experiment with more than one independent effect. variables are efficient and provide more – When there is a very small number of subjects. information than experiments with one independent variable. [ BETWEEN-SUBJECTS DESIGNS – Factorial designs – designs with two or more (1 IV, MULTIPLE GROUPS) ] independent variables Sometimes, it takes more than two treatment – Independent variables are called factors Two conditions to make a good test of a hypothesis factor experiment – the simplest factorial design. When amount or degree of the IV is important – Different values of the same variable can Factorial Designs produce different effects. – They give us information about the effects of – May give complete understanding of how the each independent variable in the experiment – variable works. main effects. Each group is run through a different condition – They enable us to answer the question: How One treatment condition may be the control does the influence of one independent variable Usual multiple independent groups affect the influence of the other variable? The Main Effects – 2 (name type) x 2 (name length) – The action of a single independent variable in between-subjects factorial design an experiment. 2. Factor and Levels Method – A change in behavior associated with a change – 2 x 2 (Type of Name: given, nickname x in the value of a single independent variable or Length of Name: short, long) between-subjects factor in the experiment. factorial design. – There are as many main effects as there are – 2 (given name or nick name) x 2 (short or long factors. name) between-subjects factorial design. Looking for Interactions ✣ 2 x 3 x 2 factorial design – Factorial design allows us to test for the – There are 3 numbers which means there are 3 relationships between the effects of different factors or independent variables independent variables. – Factor 1 has two levels – An interaction is present if the effect of one – Factor 2 has three levels independent variable changes across the levels – Factor 3 has two levels of another independent variable. – There are 12 separate conditions – Alcohol alone or sleeping pill alone can reduce anxiety, together, they can be fatal. Choosing A Between Subjects Design – The effects of one factor will change Practical reasons for keeping factorial designs depending of the level of the other. simple: – An interaction can tell us that there are limits – More treatment condition means more to the effect of one or more factors subjects. – Two independent variable = one interaction – More treatment condition means more time to – More than two independent variables result to run the experiment. more complex interactions – higher-order – More treatment condition means more time to interactions. do the statistical analysis. – It is possible to have interactions but no main – Complicated design are virtually effects or uninterpretable. – Significant effect with no interaction – Four way interactions are practically impossible to conceptualize and explain. Laying Out Factorial Design – 2 x 2 factorial design has 3 possible effects. Step 1: Indicate the 2 independent variables (ex. – 2 x 2 x 2 factorial design has 7 possible given name + nickname) effects. Step 2: Indicate the levels of factor 1 (ex. type of name: given name + nickname) 10. Within Subjects Design Step 3: Indicate the levels of factor 2 (ex. length of name: short + long) ✣ Within Subjects Experimental Design - Step 4: Indicate the 4 treatment conditions (ex. compares two or more different treatment short given name, short nickname, long given conditions (or compares treatment and control) name, long nickname) by observing or measuring the same group of individuals in all of the treatment conditions DESCRIBING THE DESIGN being compared. ✣ 2 x 2 (two by two) factorial design – A within-subjects design looks for differences Other methods of describing design variables: between treatment conditions within the same 1. Factor-Labeling Method group of participants. – 2 x 2 (Type of Name x Length of Name) – A within subjects design is often called a between-subjects factorial design. repeated-measures design because the research study repeats measurements of the same – progressive error: changes in participant's individuals under different conditions behavior or performance that are related to – It is used in experimental situations comparing experience over time in a research study but not different treatment conditions and also to related to a specific treatment or treatments (e.g. investigate changes occurring over time. practice effects and fatigue). Advantages Dealing With Time-Related Threats And Order It requires relatively few participants. Effects It essentially eliminated all of the problems controlling time based on individual differences that are the – if the different treatment conditions are primary concern of a between-subjects designs scheduled over a period of weeks, the chances – a within-subjects design has no differences greatly increase that the results will be between groups. influenced by some outside event (history) or – each individual serves as his or her own maturation or change in the measurement control or baseline. instrument. when a within-subjects design is not a Disadvantages good idea Time-related problems – e.g. comparing two methods of teaching – participant attrition reading to first-grade children (carryover – history: any outside events that occur during effects). the time that a within-subjects experiment is counterbalancing being conducted and has an influence on the – involves changing the order in which treatment participants’ scores. conditions are administered from one participant – maturation: any physiological or to another. psychological changes that occur in a participant – the goal is to use every possible order of during the time a within-subjects experiment is treatment with an equal number of individuals conducted and that can influence the participating in each sequence. participants’ scores (e.g. young children). – the purpose of counterbalancing is to eliminate – instrumentation: refers to changes in the the potential for confounding by disrupting any measurement instrument that occur over time systematic effects from factors related to time or (e.g. observer changes) the order of treatments. – statistical regression: or regression toward the – e.g. with two treatments one half of the mean is a mathematical phenomenon in which participants begins in treatment 1, then moves to extreme scores (high and low) on one treatment 2 and the other half begins in measurement tend to be less extreme on a treatment 2, then receives treatment 1. second measurement (especially a problem when participants are selected for their extreme Easy Case scores). – Order effects evenly distributed between the order effects treatment conditions. – carryover effects: changes in behavior or – It doesn’t matter which treatment comes first. performance that are caused by participation in – There is a constant (e.g., d=5 points) change an earlier treatment condition due to order effect. – carryover effects exist whenever one treatment condition produces a change in the participants Limitations Of Counterbalancing that affects their scores in subsequent treatment – Counterbalancing balances the effect of conditions (e.g. new skill from treatment 1 can ordering effect but it doesn’t eliminate it. So influence results in treatment 2). both means are inflated. – A more serious problem is when If the data are measured on an ordinal scale ( or counterbalancing adds order effect to some of can be rank ordered), a Wilcoxon test can be the individuals but not to all. used to evaluate significant differences. – When the order effect is not symmetrical. One If the data includes only positive and negative treatment might produce a larger order effect (nominal) effects then we use a sign test. than the other treatment – (Math & Statistics) Matched- Subjects Designs – In such situations, the order effects are In a matched- subjects design, each individual not symmetrical, and counterbalancing the order in one group is matched with a participant in of treatments does not balance the order effects. each of the other groups. – Number of treatments. With only two The goal of a matched- subjects design is to treatment conditions, complete counterbalancing duplicate all the advantages of within- and is easy: There are only two possible sequences. between- subjects designs without the However, as the number of treatments increases, disadvantages of either one. complete counterbalancing becomes more complex. 11. Within-Subjects Design - Small N Partial Counter-Balancing SMALL-N DESIGNS AND SOLVING One solution to this problem is to use what is PROBLEMS (CONTROL ISSUES) known as partial counter-balancing. Within Subjects Designs (Small N) A simple and unbiased procedure for selecting N stands for the number of subjects in an sequences is to construct a Latin square. experiment The designs we have discussed in the course so ✣ Latin Square - a matrix of n elements far have had many participants per group (letters) where each element appears exactly In small-N designs you can do experiments once in each column and in each row. with very few participants. – The Latin square is not a perfect example of – in fact you can do studies with only 1 partial counterbalancing because it does not participant. balance every possible sequence of treatment B.F. Skinner was an advocate of small-N conditions. For example, the first three groups designs. all receive treatment A followed immediately by – in fact he was a critic of large-N designs treatment B. – Skinner advocated eliminating as much error variance as possible ✣ Random Order - one method for improving Two main sources of error variance the Latin square is to use a random process to 1.) that do to individual differences among rearrange the columns (for example, a coin toss participants to decide whether or not each column is moved). 2.) That due to ineffective control procedures If you only use one subject in your experiment Statistical Analyses individual differences will be eliminated as a With two treatment conditions, source of error variance a repeated- measures t test If you have highly controlled experimental For more than 2 treatments a single-factor procedures that will help with physical ANOVA ( repeated measures) can be used to extraneous variables. evaluate the statistical significance of the mean Operant chamber: no light enters, sound difference. attenuated. Large N designs can obscure actual Ordinal & Nominal Scale effects  Who uses Small–N designs Variations of ABA: you can do it more than 1.) clinical psychopathology once (ex: the ABABA design with the husband – small numbers of patients and wife leaving clothes in the living room). – can ensure everyone gets treatment You can also extend the conditions in a small 2.) Animal researchers N experiment (ex: ABACADA where B, C, and D SMALL–N DESIGNS represent 3 different treatment conditions) used for practical reasons; ex: if researchers Small N design is often used to test the effects want to study animal brains or tissue, the animal of positive or negative reinforcement on has to be sacrificed --> makes more sense to use individuals with behavioral problems; ex: as few subjects as possible in cases like this sometimes rewards for positive or improved most often used in experimentation with behavior was rewarded and negative behavior operant conditioning; also used in clinical was punished which caused changes in the IV. psychopathology and psychophysics (how we Multiple baseline design: a series of baselines sense and perceive physical stimuli) and treatments are compared within the same B.F. Skinner studied positive and negative person, but once a treatment is established, it is reinforcement --> “the experimental analysis of not withdrawn. behavior” states it is better to use careful, — used when it is not desirable to reverse continuous measurements rather than statistical treatment conditions, when the researcher wants tests. to test a treatment across multiple settings, or Baseline - a measure of behavior as it normally when the researcher wants to assess the effects occurs without the experimental manipulation. of a treatment on several behaviors. to make sure maturation is not a confounding Statistics are not usually used in small n threat, small n designs remove the IV and return designs; you can look at the data to see changes to the original control condition after completing in the IV rather than running statistical tests. the experimentation manipulation (goes back to — the role statistics do play is to infer things measuring the plant every Monday for three about the population from sample data because months without talking to it) making generalizations based on one subject isn’t reasonable. Types of Small N designs When to use/advantages of small N designs: ABA design — when you are studying a particular subject — A baseline (ex: a disturbed child) — B treatment — when very few subjects are available — A return to baseline — you get a more accurate picture of results or Rat bar pressing effects (because you measure the effects — 1 hour no Rf multiple times and observe it closer) — 1 hour Rf — low in external validity; it may be hard to — 1 hour no Rf generalize for the entire population based on the ABA Designs: refers to the order of the results of a few subjects; EX: reactions may be conditions of the experiment; A (baseline different for individuals in different situations condition) followed by B (experimental (ex: we all get startled at a loud gunshot noise, condition) returns back to A but some people may laugh with relief that it — may be used only if the treatment conditions was the researcher that shot the gun or be angry are reversible. about it — also called reversal designs — history threats are a problem with small N — can be used for large N designs too designs so it is important to replicate findings before generalizing them; ex: someone nice could come give the plant fertilizer right as 2. Check assumptions to ensure your data is researchers begin talking to it. satisfactory for performing the inferential statistic (or choosing the correct statistic 12. Psychological Statistics in Experimental depending on which assumptions are met). Research 3. Perform the test by running the inferential statistic ✣ Statistics - the branch of statistics that 4. Interpret the results and make a decision about involves summarizing, tabulating, organizing, whether you reject or fail to reject the null and graphing data in order to summarize a data hypothesis, write-up the results in APA format, set for the entire population of subjects. No and provide a visualization of the results. attempt is made to infer or predict the characteristics of subjects that have not been Null Hypothesis measured or observed. Symbol: H0 H0 = IV had no effect on DV Alternative Hypothesis Symbol: H1 H1 = IV had an effect on DV We either reject the null hypothesis or fail to reject the null hypothesis. ✣ Directional Hypotheses - also called Two Different Types of Statistics one-tailed hypothesis because only one tail of the distribution would lead us to fail to reject the ✣ Descriptive statistics are used to summarize, null hypothesis. One direction of the effect. organize, and overall describe our sample data. Typically, we do so using measures of central ✣ Non-directional Hypotheses - would not tendency (e.g., mean, median, mode), measures know whether the difference will be greater or of dispersion (e.g., range, standard deviation, less than 0, but there will be a difference; these variance), and shape (e.g., skew, kurtosis). We are also called two-tailed hypothesis because may also visualize the data using tables or both tails of the distribution would lead us to fail graphs. to reject the null hypothesis. ✣ Inferential statistics are what we use when we collect data about a sample and see how well that sample infers things about the population from which the sample comes from. Typically, we do so with statistical tests like the t-test, ANOVA, correlation, chi-square, regression, and more. ✣ Hypothesis Testing - allows researchers to use sample data to draw inferences about the population of interest. It is based on chance or probability (aka our p-value). 1. Look at the data by examining the descriptive statistics and describing your hypotheses. ✣ T-Test for Independent Samples or 2. Description of your data. If you fail to meet Independent T-Test assumptions, you should specify that and The independent t-test is used to test the describe what test you chose to perform as a difference in the dependent variable between result. two different groups of observations. 3. The results of the inferential test, including The grouping variable is the independent what test was performed, the test value and variable. degrees of freedom, p-value, and effect size. In other words, we use the independent t-test 4. Interpretation of the results, including any when we have a research question with a other information as needed. continuous dependent variable and a categorical independent variable with two categories in The research question was whether there was a which different participants are in each category. difference in student grades between Spearman’s The independent t-test is also called the and Skinner’s classes. Spearman’s students (M = independent samples t-test and the Student’s 75.1, SD = 8.83, n = 15) had significantly t-test. higher grades than Skinner’s students (M = There are three different types of alternative 69.5, SD = 4.87, n = 15), t (31) = 2.15, p = hypotheses we could have for the independent.040, d =.74. t-test: 1. Two-tailed ✣ T-Test for Dependent Samples or Matched H1: Group 1 has a different mean than Group Groups 2. The dependent t-test is used to test the H0: There is no difference in means between difference in the dependent variable between the two groups. two categories in which participants are the 2. One-tailed same across categories. H1: Group 1 has a greater mean than Group 2. The category variable is the independent H0: The mean for Group 1 is less than or equal variable. to the mean for Group 2. In other words, we use the dependent t-test 3. One-tailed when we have a research question with a H1: Group 1 has a smaller mean than Group 2. continuous dependent variable and a categorical H0: The mean for Group 1 is greater than or independent variable with two categories in equal to the mean for Group 2. which the same participants are in each category. Assumptions Assumptions 1. The dependent variable is normally 1. The differences in scores in the dependent distributed variable are normally distributed 2. Variances in the two groups are roughly equal 2. The dependent variable is interval or ratio (i.e., homogeneity of variances) (i.e., continuous) 3. The dependent variable is interval or ratio 3. Scores are independent across participants (i.e., continuous) 4. Scores are independent between groups Decide which statistical test you should be using — If you violated the assumption of normality, and no transformation fixed your data, then you can perform the non-parametric version of the Independent T-Test - APA Write-up Format dependent t-test called the Wilcoxon rank. As a 1. Description of your research question and/or reminder, non-parametric tests do not make hypotheses. assumptions about the distribution