Experimental Psychology: Chapter 5

• Chapter 5 discusses correlational and quasi-experimental designs in experimental psychology.

Quasi-Experiments vs. Actual Experiments

• Quasi-experiments resemble real experiments, but without manipulating antecedent conditions and/or random assignment.
• These studies look at effects of pre-existing conditions or subject characteristics on behavior.
• For instance, an examination the incidence of Alzheimer's in ibuprofen users versus non-users after age 50 could be a quasi-experiment.
• Experiments involve researchers assigning subjects to conditions they create.
• Quasi-experiments are suitable when antecedent conditions should not or cannot be manipulated.
• Studying the impact of spousal abuse on child abuse frequency is an example where quasi-experiments are appropriate.

Properties of Correlation

• Pearson correlation coefficients assess simple correlations, shown as r (50) = +.70, p = .001.
• Correlation coefficients are described by linearity, sign, magnitude, and probability.
• Linearity indicates if the relationship plots as a line or a curve.
• Sign indicates if the correlation is positive or negative.
• Magnitude measures the strength, ranging from -1 to +1.
• Probability denotes the likelihood of obtaining the observed magnitude by chance.

Scatterplots

• Scatterplots are a graphic display of data point pairs on x and y axes.
• Scatterplots depict correlation's linearity, sign, magnitude, and probability.

Range Truncation

• Range truncation artificially restricts X and Y ranges, reducing correlation strength.

Outliers

• Outliers are extreme scores affecting data trends and correlations
• Range truncation eliminates outliers.

Coefficient of Determination

• The coefficient of determination (r²) estimates a predictor variable's variability amount.
• Handshake firmness accounted for 31% of first impression positivity.

Correlation vs. Causation

• Correlational studies do not establish causation since there are no manipulated independent variables or random assignment.
• Three reasons correlations cannot prove causation:
• Casual direction
• Bidirectional causation
• The third variable problem
• Causal direction occurs because correlation is symmetrical, and B may as easily cause A as A causes B.
• Bidirectional causation is where two variables affect one another.
• The third variable problem occurs when a third variable creates the appearance of a relationship between the other two.

Multiple Correlation (R)

• Multiple correlation (R) is used to determine relationships among three or more variables.
• Age, TV watching, and vocabulary were measured, with an R of +.61.

Partial Correlation

• Partial correlation involves keeping one variable constant to assess its influence on the correlation between two others.
• Age can be held constant to measure how television viewing affects vocabulary.

Multiple Regression

• Multiple regression predicts behavior using scores from multiple variables.
• Estimating vocabulary by inputting age and TV watching habits as predictor variables.

Causal Modeling

• Causal modeling creates and tests models suggesting cause-and-effect relationships.
• Path analysis and cross-lagged panel designs are forms of causal modeling.

Path Analysis

• Path analysis creates and test models of causal sequences via multiple regression.

Cross-Lagged Panel Design

• In cross-lagged panel design, relationships measured over time suggest a causal path.

Ex Post Facto Design

• Ex post facto designs examine effects of pre-existing subject variables without manipulating them.

Non-Equivalent Groups Design

• Non-equivalent groups design compares treatment effects on pre-existing groups.
• In Company A fluorescent lighting is installed, productivity is then compared to incandescent lighting in Company B.

Longitudinal vs. Cross-Sectional Approaches

• Longitudinal study measures the same subjects at different times to see the effect of time.
• Cross-sectional studies compare different developmental stages or classes simultaneously.

Pretest-Posttest Design

• Pretest/posttest designs measure behavior before and after an event.
• This is said to be quasi-experimental, as there is no control condition.
• Practice GRE test 1, six-week prep course, Practice GRE test 2 exemplifies this.

Internal Validity Problems with Pretest-Posttest Design

• Lacks a control group that receives a different IV level or no preparation course.
• Practice effects (pretest sensitization) may confound the results due to reduced anxiety and learning from pretest answers.

Solomon 4-Group Design

• The Solomon 4-group design includes these conditions:
• Receives pretest, treatment, and posttest
• Receives pretest and posttest only
• Receives treatment and posttest only
• Receives posttest only

Hypothesis Definition

• Hypotheses explain relationships between two or more variables.

Experimental Hypotheses

• Experimental hypotheses tentatively explain an event or a behavior.
• It predicts the impact of an independent variable on a dependent variable.
• Cognitive behavior therapy (CBT) produces less relapse than antidepressants, for instance.

Non-Experimental Hypotheses

• Nonexperimental hypotheses predict how variables might correlate, but not establish causation.
• Red-haired patients receive less pain relief from medication than blonde patients is an example.

Synthetic Statements

• A hypothesis must be a synthetic statement, capable of being true or false.
• Examining the effects of morning meals on student reading ability is an example.
• "Hungry students read slowly."

Testability

• An experimental hypothesis is testable via manipulating an IV and measuring the DV results.

Parsimony

• Parsimoniousness means a simpler hypothesis is preferred over a complex one.

Intuition

• Intuition should be guided by literature review.

Helpful Strategies for Developing Hypotheses

• Helpful strategies: read a psychology journal, observe people in public, and identify real-world problems with causes.

APA-Format Paper Introduction

• The introduction selects relevant research related to the hypothesis.
• It shows how the study furthers knowledge by addressing unanswered questions.

Value of Meta-Analysis

• Provides helpful information on the topic.
• It is not an experiment, however it is a statistical analysis that includes many studies.
• Measures the average effect size of an independent variable across studies with similar methods.
• Establishes strength and external validity of causal relationships.

Independent Variables

• Independent variable (IV) is the variable or antecedent condition intentionally manipulated by an experimenter.
• Independent variable levels are experimenter-created values of the IV.
• Experiments require at least two levels.

Confounding Explained

• An experiment is confounded when an extraneous variable value changes with the independent variable.
• Experimental subjects in the morning and control subjects at night exemplify this.

Dependent Variables

• Dependent variables measure outcome; the experimenter measures the change in behavior produced by the independent variable.
• The value of the dependent variable depends on the independent variable value.

Operational Definitions

• Operational definitions specify a variable's experimental meaning.
• It is defined in terms of observable operations, procedures, and measurements.

Experimental Operational Definitions

• Experimental operational definitions specify the procedure to create independent variable values.

Measured Operational Definitions

• Measured operational definitions specify the procedure to measure the dependent variable.

Types of Scales

• Nominal scales assign items to categories with shared features, without measuring magnitude.
• Sorting animals into friendly and shy categories.
• Ordinal scales measure magnitude using ranks without precise values.
• Interval scales measure magnitude with equal intervals but lack absolute zero.
  • Degrees Celsius or Fahrenheit.
  • Sarnoff and Zimbardo's (1961) 0-100 scale.
• Ratio scales measure magnitude with absolute zero and equal intervals.
  • Distance in meters or time in seconds.

Reliability

• Reliability indicates consistency of experimental/measured operational definitions.
• An accurate bathroom scale is an example.

Interrater Reliability

• Interrater reliability measures how much observers agree when measuring behavior.
• Example: agreement among three observers when scoring personal essays for optimism.

Test-Retest Reliability

• Test-retest reliability is how consistent a person's scores are across multiple administrations.
• Wechsler Adult Intelligence Scale-Revised exhibits highly correlated scores when administered twice, two weeks apart.

Interitem Reliability

• Interitem reliability indicates consistency across different instrument parts that measure the same variable.

Validity

• Validity means an operational definition accurately manipulates the independent variable or measures the dependent variable.

Face Validity

• Face validity judges if a manipulation/measurement technique has self-evident validity.
• Using a ruler to measure pupil size.

Content Validity

• Content validity indicates how well a measurement samples the content of the dependent variable.
• When an exam only contains questions on chapter 2 when the exam is supposed to be over chapters 1-4, it has poor content validity.

Predictive Validity

• Predictive validity measures how well a procedure predicts future performance.
• ACT scores correlating with college GPA.

Construct Validity

• Construct validity is how well an operational definition represents a construct.
• Parent constructs that include perceptions of others as unfriendly.

Internal Validity

• Internal validity means the changes in the dependent variable came from the experimental conditions.
• Establishes a cause-and-effect relationship.

Confounding

• This occurs when extraneous conditions systematically change across the experimental conditions.
• Studying meditation and prayer effects on blood pressure becomes confounded if one group exercises more.

History Threat

• Occurs when outside events threaten validity via altering the dependent variable.
• Measuring group A before lunch and group B after lunch.

Maturation Threat

• Comes when subjects' physical or psychological transformations threaten validity by changing the DV.
• Boredom increasing subject errors.

Testing Threat

• Threat comes when exposure to the testing affects performance.

Instrumentation Threat

• Instrumentation threat occurs when changes in the measurement instrument affect internal validity.
• Reaction time that becomes less accurate.

Statistical Regression Threat

• Occurs when conditions are assigned via extreme scores, procedure is unreliable, and subjects are retested.

Selection Threat

• Selection threat occurs when individuals are assigned to conditions by experimental assignments.

Subject Mortality Threat

• Subject mortality happens when subjects drop from experimental conditions at different rates.

Selection Interactions

• Selection threats interact with history, maturation, statistical regression, subject mortality, or testing.

Method Section Purpose

• Method details experiment’s Participants, Apparatus/Materials, and Procedure.
• Readers get sufficient detail to exactly replicate the experiment.

When to Use Apparatus Section

• Apparatus is needed with unique specialized equipment, or when capabilities of common equipment must be explained.

Physical Variables

• Day of week, experimental room, and lighting are aspects of the situation that should be controlled.

Elimination

• Elimination fully removes extraneous physical variables from the situation.

Constancy of Conditions

• Constancy controls extraneous physical variables by keeping treatment conditions identical, except the IV.

Balancing

• Balancing distributes extraneous physical variable effects across treatment conditions.
• It reduces the impact of variables that cannot be eliminated.

Ordered Techniques

• Eliminate, keep constant where elimination is not possible, or balance social conditions where constancy is not possible.

Social Variables

• Social variables that influence experimental results include demand characteristics and experimenter bias.

Demand Characteristics Explained

• Demand characteristics are the experimental situation cues that elicit specific participant responses.
• Students try and cue professors to end the lectures by packing up.

Threat to Internal Validity

• Demand characteristics can confound experiments by varying across experimental conditions.
• Subjects may even act to try and confirm the hypothesis.

Single-Blind Experiments

• These do not tell subjects the treatment condition.
• A single-blind drug study capsules look and taste identical.

Why Use Single-Blind Experiments

• Treatment conditions eliminate cues that might alter behavior.

Placebo Effect

• Placebo occurs when a participant improves from a treatment they received.

Demand Characteristics Controlled

• Cover stories uses false or plausible explanations to disguise what the hypothesis is.
• The explanation must be used scarcely sincethey are form of deception.

Experimenter Bias

• Any behavior from the experimenter that could confound the experiment.
• Researchers can also provide more attention to subjects in one of the conditions than in another.

Rosenthal Effect

• Rosenthal Effect occurs when experimenters who treat subjects based on their resulting influenced subject performance
• It is the Pygmalion effect of self-fulfilling prophecy.
• By paying more attention and giving more feedback to high aptitude student is a key example.

Minimize Experimenter Bias

• Since subjects are blinded, single-blind designs only control single bias.
• Double-blind experiments control demand characteristics and experimenter bias because the subjects and experimenter are blinded

How Personality Affects Experimental Results

• Warm and friendly equals more pleasant results.
• If they are hostile this means there will be an inferior subject performance.

Control Personality Variables

• Multiple experimenters balance the amount of the test subjects.
• Statistical experimenters are used with independent variables.
• There must an interaction that has been confounded with single experimenters to the follow contact minimizes.

Volunteers vs. Non-Volunteers

• Non-volunteers score poorly while volunteers sore higher
• Political attitudes are highly regarded and are less authoritarian.

Context Variables

• There are variables that score less on intelligence tests from nonvolunteers.

Select the Experiment

• Allows the subjects to sign up to get the results of the heavy metal experiment.

Folklore Summary

• Select the experiment that relies to depend on the folklore about the subjects.
• It can affect the students results that have similar signs.

Experimental Design Purpose

• To test the hypothesis, an experimental design experiments a plan.
• You can apply designs to investigation even if there are different hypotheses.

Key Factors

• Hypothesis for factors is determined by experimental design from three factors.
• Independent variable that has a number
• All results are subjective from experimental conditions

Between-Subjects Design

• Only participates if a subject in between-subject designs.

Generalize the Findings

• Sample determinations can generalized results from population.
• Validity increases when sampling at random.

Group Minimum Subjects

• Subjects has its conditions should be detected as a treatment with effects.

Effect Size Importance

• Treatment with effects is estimated statistically.
• Relationships could get stronger if the variables are independent.

Sample Sizes and Effects Size

• Determines what factors require to experiments, by requiring researchers or programs.

Two-Group Design

• Involves two or more separate subjects.
• Includes versions of this design that can be controlled.

Independent Designs for Research

• Subjects are randomly assigned with multiple levels for one of the conditions.
• Two and independent groups were designed with control groups.

Random Assignment

• Equalizing conditions the can experiment to each other will prevent the study from being controlled.

Conditions Differ

• Independent conditions can present values with variable tests.
• One level receives the same procedure while variable independent has a zero
• Independent can experimental a treatment or procedure.

Two Designs Discussed

• This experiment can assign multiple level with multiple variables.
• Designs requires that the right variables be distributed.

Assign Subjects

• Conditions for effectiveness limit assign what's needed.
• Multiple people control it by their variables.

Two Groups Discussed

• If conditions are determined randomize the what can occur.

Matching Explained

• Pairs that can be formed with identical forms.

Two Matched Test

• Can be randomly assigned with the right conditions.

Block Randomization

• Equal multiple are assigned from randomly.
• Conditions are randomly assigned to a assigned test with determined blocks.

Researcher

• Pilot studies or researchers can pre determine what needs to be done.

Experimental

• Number of treat,nets will gain from extra results.

Practical Limitations

• Test that will determine what can be promising.

Pilot Study

• Trials that can determine if an experiment may work.
• Multiple experiments may be refined if there is help.