Quantitative Psychology 2 PDF
Document Details
Uploaded by InestimableNovaculite2278
Tags
Summary
This document provides an overview of quantitative methods in psychology, specifically focusing on causality and its relationship to correlation.
Full Transcript
EEK 10 W Quantitative Methods 1: Causality ypes of research: descriptive; relational; comparative (experimental and T quasi-experimental) A brief history of research methods: - s early as 1909, Robert Sessions Woodworth was giving his students copies of a A...
EEK 10 W Quantitative Methods 1: Causality ypes of research: descriptive; relational; comparative (experimental and T quasi-experimental) A brief history of research methods: - s early as 1909, Robert Sessions Woodworth was giving his students copies of a A mimeographed handout called “Problems and Methods in Psychology” - When it was finally published in book form asExperimentalPsychologyin 1938, the publishers’ announcement merely said, ‘The Bible is out’. The purpose of the experiment - is to demonstratecausality. … - For this to happen, there are 3 requirements: - Cause must precede effect in time-temporal precedence - Cause and effect must be empirically correlated with one another-covariation - The relationship between cause and effect cannot be explained in terms of a third variable-eliminated 3rd variables - Should the cause be bothnecessary and sufficientto explain the effect? Do you have to have the cause to get the effect, and is only the cause enough? (e.g. frustration causing aggression) Causality and correlation CORRELATION DOESN'T EQUAL CAUSALITY - oes chocolate reduce cognitive decline? Chocolate was found to increase cognitive D functioning in rats. Study mapped number of Nobel prizes against chocolate per capita in various countries. - Does this meet the criteria for causality? EXPERIMENTAL DESIGNS What are features of an experiment? - Establishing independent variables - Creating experimental conditions or comparisons that are under the direct control of the researcher - Controlling extraneous variables - Not of interest to the researcher - Must be controlled, otherwise it leads to confounding - Measuring dependent variables - Behaviours that are measured in the study - Must be defined precisely . Independent variables 1 - Must have a minimum of two levels - May be: - Manipulated (something you change) or subject variables (age) - manipulated: - Situational– features of the environment that participantsmay encounter (time of lecture) (e.g. noise or silence) - Task– variations in the task given to participants - Instructional– variations in the instructions given- eg: verbal or written . C 2 ontrolling for extraneous variables - We need to control anything that’s not of interest to us (i.e., we need to eliminate any “third variable” explanations) .g. experimenter effect: when the researcher gives subtle clues about the way in e which they expect the subjects of a study to respond - emand characteristics: certain features of the research setting that can impact the D findings. What's the confound? drug company developed a new medication to control the manic phase of bipolar A disorder. The firm hired a hospital psychiatrist to test the effectiveness of the drug. He identified a group of manic-depressive patients and randomly assigned them to a drug or placebo group. Nurse Smith was told to administer the drug and Nurse Johnson was told to administer the placebo. Each nurse made daily observations of her patients during treatment. A month later the observations were compared. In general, patients in the drug group had behaved more "normally" than patients in the placebo group. The drug company publicised its product's effectiveness. . M 3 easuring the dependent variables - The credibility of an experiment depends on (amongst other things) the operational definition of the measured outcomes - Operational definition = how the construct is defined in terms of the specific operations, measurement instruments, or procedures through which it can be observed EXAMPLE: Bandura & Ross study on a blow up doll xperimental group 1: “real-life” aggression E Experimental group 2: human film aggression Experimental group 3: cartoon film aggression Control group: no exposure to aggressive models - Looking at the effects of exposure to violence on aggression - Operationally defined as what the child would do to the blow up doll Validity - tatistical conclusion validity S - Construct validity - External validity - Internal validity . S 1 tatistical conclusion validity Do the correct analysis, without violating any assumptions- normality, homogeneity of variance, independence Report all analyses, even the ones you don’t like Don’t go fishing (Type I error – the more the analyses, the higher the chance of a false positive) Make sure your measures are reliable, so that if there is an effect you can find it (otherwise you run the risk of Type II error – a false negative) . C 2 onstruct validity Definition: the extent to which a measure of a construct isempirically relatedto other measureswith which it istheoretically associated Does a test trulymeasure what it purports tomeasure?Are youroperational definitionsadequate? - Discriminant evidence: measures that aren’t theoreticallyrelated aren’t empirically related either - Convergent evidenceis the opposite 3. External validity T he degree to which the findings / conclusions can be generalizedbeyond the confinesof thedesign and study setting. eneralisability to: wont always generalize to all 3, but if you can great G - Other populations (e.g. males/females): the sampling can limit the study’s generalizability - Other environments (many studies are performed in labs). Ecological validity: can the findings be generalised to real-life situations? - Other times . Internal validity 4 The degree to which a study ismethodologically soundandconfound-free (designed well; no third variables) Degree to which we can be sure that the dependent variable changed as a result of the independent variable. The findings follow directly and unproblematically from the research design Concerns the research design and validity of the conclusions. re-Post Studies:the dependent variable is measured;there’s an intervention; the P 0 x 0 0 p 0 dependent variable is measured again. est anxiety could also be decreasing through natural development (getting used to it). Only T choosing 1styears means this is likely to be thecase (selection bias). Can’t assume that all changes of the DV are due to the intervention: NB: Threats to internal validity: Know how to define each ,and how we would remedy or counter it - istory:anything else that happensduringthe studythat could affect the outcome H of the study (e.g. the tests were changed, the participants watched something that changed them changes the DV) something influencing the results between pre and post test - Maturation:natural development/independent changein the participants (getting better or worse at something) - Testing: reactive effects to participatingin thestudy. Theinitial test/measurement maychange the DVor how the participants behave.The experimenter may suggest something that increases or decreases the DV. Includes priming, demand characteristics and experimenter effects. - Instrumentation: the measurement / test isunreliableor problematicin some way - Selection bias:non-randomgroupassignment: groupsare different in an important way. Experimental group and control group shouldbethe same. Need to get a large enough sample, then randomise (most relevant in experimental groups) RALS - Attrition (mortality):non-randomparticipantdrop-out.They may share particular attributes, thusskewing the results. - Regression to the mean: scorestend to average out.Data that isinitially extremely higher or lowerthan the mean will likelybe closer to the mean if it is - Instruments: some people may lie to be liked , double tests. example stress survey then do electrodes to see heart. - Larger sample size is better, might just regress to the mean does not show the treatment is working. Standardized everything- don’t want to think the treatment is working if its not because people are regressing to the men (false positive) measured a second time (or vice versa) - To counter it: standardization of measurements if one group gets therapy ensure all get therapy. Random assignment , if large enough sample size RALS Control problems xperimental design: strongest internal validity (history, maturation, testing effect, selection E bias are controlled for) Instrumentation effect is also controlled for unless the measurement/test varies significantly between measures Regression to the mean is controlled for if the two groups were drawn from the same sample Attrition can still be an issue Try to make sure that the only difference between the two groups in an Two basic experimental designs: experiment is the difference you as the researcher provide . Between-subject designs 1 2. Within-subjects (repeated measures): - Between-subjects:each person takes part in only onecondition of the research - Problem: creating equivalent groups - Within-subjects(repeated measures): each person takespart in all conditions of the research - Problem: participating in one condition might affect behaviour in another condition (sequencing effects) ry to make sure that the only difference between the two groups in an experiment is the T difference you as the researcher provide Between subjects designs - KA independent measures. A - Each person takes part inonly one conditionof theresearch - Most commonly used type - Used when theIV is a subject variable(e.g., gender;extrovert/introvert; marital status) -Also used whenexperience gained in one levelwouldmake it impossibleto participate inanotherlevel Problem: making the groups equivalent Two ways to try to have equivalent groups: Random assignment and matching a. Random Assignment / randomisation - this will prevent any systematic bias - The primary purpose of random assignment is to control for potential confounding variables or biases that might affect the internal validity of the study. By randomising participants to groups, researchers aim to distribute both known and unknown variables evenly across those groups, making them statistically comparable - Goal: to takeindividual factorsthatcould biasthestudy, andspread them evenly throughout the different groups - Randomness isn’t a guarantee Works best withlargenumbers b. Matching (nota true experimental design) - ot an experimental design because the groups are no longer random: opens up N room for bias. Not ideal, because theremay be otherfactors effecting the DVthat aren’t recognised useful only small no. of participants available and cant RA(LS) - Matching participants across groups whoshare similarfactorsthatcorrelates with and is thereforeexpected to affect the DV - The main purpose of matched-group design is toenhancethe comparability of groupsby ensuring that they are matched on specificvariables that might influence the dependent variable. This can reduce the potential for confounding variables and increase the internal validity of the study. - because each of the match participants goes into a different group defect of the characteristic on which the participants were matched gets distributed evenly across treatments - as a result this characteristic contribute a little to the differences between groups - The error contributed by this characteristic has therefore been minimised - Choose a matching variable that correlates with the DV (i.e., is expected to affect the outcome in some way) - Make sure there is some reasonable way ofmeasuringparticipants on the matching variable. Don’t match people on something when it has nothing to do with your study – e.g. height .g. comparing reading skills between people with and without brain lesions: match e participants according to education etc (things that may affect the DV) advantage - an advantage over matching over random assignment is that it allows you to control subject variables that may otherwise obscure the effect of the independent variable - Useful when you only have asmall numberof participantsavailable (e.g. if the phenomena is rare), and/orcan’t use random assignment. disadvantage - What happens if the match characteristic does not have the effect on the dependent variable that you may initially thought it might? ○ Control in between subjects designs - between subjects designs must be carefully controlled - there are two main issues when considering control in an in between subjects design articipant variables p - each participant gets assigned to one condition only - participant mood may affect the experiment - this is the error ontrolling for this : c Matched-group design and random assignment are two different methods used in experimental research to create comparable groups or conditions. While they serve similar purposes, they are distinct approaches, and they can be used in conjunction to enhance the validity of an experimental study. Within subjects designs ithin-subject designs, also known as repeated-measures designs, are a type of W experimental design used in research to study the effects of one or more independent variables on a single group of participants. roblem: participating in one condition mightaffectbehaviour in anothercondition P (order/sequencing effects) Advantages: - eed fewer people N - No problems with equivalent groups - Reduces error variance, since you have no between-condition individual difference – so it gives more statistical power to find an effect if there is one Problems – sequence or order: - ractice effects - The influence taking part in one condition has on subsequent P performance in other conditions - can come from a number of sources: - Fatigue - habituation - Learning- if you learn how to perform a task in the first condition performance is likely to be better in a similar condition - Carryover effects (does it matter if condition A comes before condition B?) Try to fix using counterbalancing. - ssigning the various treatments of the experiment in a different order for different a participants 1. complete counterbalancing provides every possible ordering of treatments and assigns at least one participant to each ordering 2. Partial counterbalancing includes only some of the possible orders Using equivalent measures practice effects may be reduced a little: using different measures, do the BDI then do another psychological test Fatigue- break in-between WEEK 10 Single-factor designs ost research is for the scientific enterprise as an inherently nomothetic approach seeking m to establish general principles and generalisations that apply across individuals - however these general principles do not always apply to everyone s a result researchers have argued that anomotheticapproach should be accompanied a with an idiographic approach - idiographic- research seeks to describe, analyseand compare behaviour of individual participants - therefore scientists do not only focus on general trends but also on the unique behaviours of specific individuals ingle factor designs:These are a type of experimentaldesign that are used to investigate S the effect of a single independent variable and the dependent variable - nalyse the behaviour ofindividual participantsratherthan grouped data a - The units of analysis is not the experimental group but rather the individual participant - more than one participant to study but each participants responses are analysed separately and the data from individual participants are rarely averaged - cannot be analysed using inferential statistics-T tests and F tests Vocabulary: - “ Factor” –the independent variable - “Level” –number of “states” of that variable youare testing Example: - If you are studying test anxiety, then divide your group into “high test anxiety” and “low test anxiety”. “Test anxiety” is the factor, and “high” and “low” are the two levels (or conditions). key factors: - independent variable (factor) - can have different levels - dependent variable - experimental groups - Random assignment - control groups - measurement and data collection - statistical analysis - hypotheses- in the beginning advantages - roup experimental designs failed to handle three important research issues- error G variance, generalisability and reliability . e 1 rror variance - all data contains error variance - researchers must minimise error variance because I can mask the effects of the independent variable - Group experimental designs aim to solve error variance by : - averaging the responses of several participants to provide a more accurate estimate of the affect of the independent variable - by using groups of participants we can estimate the amount of error variance - it is however argued that because it uses more than one independent variable, it is more complex- greater potential for errors - also interaction effects - difficulty isolating effects - increased variability . g 2 eneralisability - Single participant researchers argue that the data from group designs do not permit us to identify the general effect of the independent variable - given that group averages may not represent any particular participants response attempts to generalise from an overall group results may be misleading . r eliability 3 - Group designs demonstrate the effect of the independent variable a single time and no attempt is made to determine whether the observed effect is reliable - when possible, single participant experiments replicate the effect of the independent variable in two ways: - intra participant replication: replicating the effectof the independent variable with a single participant - Interparticipant replication:studying the effectsof the independent variable on more than one participant - Single case designs hence allow the generality of one’s hypothesis to be assessed through replication on a case by case basis Data from single participant designs r esearchers who use single participant designs resist analysing their results in the forms of mean standard deviations other descriptive statistics he preferred method of presenting the data is in graphs that show the results individually T for each participant - graphic analysisalso known as visual inspection if the behavioural changes so pronounced enough to be discerned through a visual inspection of grass the research concludes that the independent variable affected the participants behaviour - unfortunately the results are not always this clear cut: - it is difficult to tell whether the independent variable cause a change in the behaviour during the treatment period or whether the observe change was random - when the independent variable was remove the participants behaviour changed but did not return to the original baseline level - Did the independent variable cause changes in behaviour? Advantage: - traightforward and simple s - because graphic analysis is a relatively insensitive way to examine data only the strongest effect will be accepted as real - this is in contrast to group data and which are very weak effects may be found to be statistically significant disadvantage: - ot sufficiently sensitive or objective as means of data n - very ambiguous - How big of an effect is big enough? Types of single factor designs - Independent groups, 1-factor xample: Blakemore and Cooper (1970) raise two week old cats in two visual environments: E one sees only horizontal stripes, others see only vertical stripes. - actor? visual environment F - Levels? vertical stripes & horizontal stripes Analysis of data? - t- test for independent groups - Matched groups, 1 factor lagrove (1996): Will sleep-deprived people be influenced by misleading questions? - there B won't be random assignment for this - actor? sleep deprivation F After sleep deprivation they were told a story and asked some questions about things that were not there, e.g. ask about a red car - Levels? 21 hours or 43 hours when there was no car in the story. - Problem: Analysis? - T-test for dependent groups - 1 factor, nonequivalent groups nepper, Obrzut & Copeland (1983): Are gifted children good at social and emotional K problem-solving, compared with average-IQ children? - actor? F - Levels? Analysis ? - t- test for independent groups Groups cannot be matched or randomly assigned here Gifted = 2 standard deviations on the right of the mean (on a normal distribution curve) Factor = IQ Levels = gifted or average IQ - Within-Subjects, 1-Factor ee & Aronson (1974): Will children shift their balance to moving visual stimuli as if their L balance has shifted? - actor? Visual Stimuli F - Levels? Forwards and Backwards Analysis? - t- test for dependent groups - Multilevel Designs: All the previous designs have included only 2 levels. dding more levels adds more sensitivity and can show the true relationship between A variables. Can be within or between subject designs Between-subjects, measuring only 2 levels: Factorial designs - actorial designs havemore than one independent variable-experimental design F - Multifactorial designs allow us to look atinteractioneffects Vocabulary: - ust say “this is a 3x2 factorial design” M - A 2-factor study with 3 levels of one factor and 2 of the other is a 3x2 design; it has 6 conditions: - closer approximationofreal-world settings,whereindependent variables don’t A exist alone. - All levels of each independent variable are combined with all levels of the other independent variables. - Simplest design: 2x2 factorial design 4 conditions 2-Factor with 3 levels - 2x3 design: a 2-factor study with 3 levels of one factor and 2 of the other. 6 separate conditions. ere we are looking at test performance. Specifically the effect of test anxiety in males and H females on test performance. So, 2 IVs here: test anxiety level, and gender. This means we have 2 factors. est anxiety has 3 levels (low, medium, high), and gender for this example has 2 levels T (male, female) lease note the 6 separate conditions. A condition is a subgroup created when we have P different factors, e.g. males with high anxiety would be one of the conditions here. Two kinds of results: Main effects - re exclusivelydue to onlyone independent variableor another(the effect of one a of your independent variables on the dependent variable) - In general, there isone main effect for every independentvariablein a study Example: Word-finding (DV) and sleep-deprivation and caffeine-deprivation (factors / IVs). 4 conditions. - Can look at effect of caffeine OR the effect of sleep DV = word finding 2 Factors (Sleep-deprivation; Caffeine-deprivation) 2 Levels (Yes) 4 Conditions Interaction effects - hen the effect of one IV variabledepends on thelevel of another IV. w - A statistical interaction occurs when the effect of one independent variable on the dependent variable changesdepending on the levelof another independent variable. - Example: Word-finding depends on sleep-deprivation and caffeine-deprivation. xample: an interaction with no main effects E Godden & Baddeley (1975): Is learningcontext-dependent?Four conditions: - earn on land – recall on land L - Learn on land – recall underwater - Learn underwater – recall on land - Learn under water – recall under water No real difference between land and underwater (i.e. no main effects). But learning nd recalling within the same environments (2 of the 4 conditions) was more a effective. - o here learning on land has no real advantage unless one has to recall on land as S well (that is the interaction). If the main effects were significant, then learning on land would be an advantage, or recalling underwater would have an advantage (which is not quite the case). In terms of inferential stats, there are no real differences between the main effects averages, i.e. the overall averages. However, we see that recall on land was better when they learned on land and recalled on land (compared to learning on land and recalling underwater). Similarly, recall was better underwater when they learned underwater. o, learning is somewhat context dependent, because there is an interaction between where S the material is learned and where it is recalled. Mixed factorialdesignwith counterbalancing Includes both independent groups (between-subjects factor) and within subjects factor. - Counterbalancingis used to fix theorder / sequencingeffect EXAMPLE:Riskind & Maddux (1993): looking at how self-efficacyrelates to fear - etween-subjects variable: High self-efficacy vs. low self-efficacy B - Within-subjects variable: “Looming” (participants saw a video of a spider either coming towards them or moving away) - DV: Fear ut the ordering mattered: whether participants saw the spider moving towards them first or B second - fixed using counterbalancing:randomly assignedpeopleto gettingeither orderof the within-subjects variable he sample would be split into two groups: experimental (A) and control (B). For example, T group 1 does ‘A’ then ‘B,’ group 2 does ‘B’ then ‘A’ this is to eliminate order effects. Although rder effects occur for each participant, because they occur equally in both groups, they o balance each other out in the results. Interaction effect between ‘looming’ and ‘low self-efficacy’ Results: Number of participants needed: - emember that the number of participants needed goes up as your number of R conditions increases: - 2x2betweensubjects: 5 per group, 20 in total - 2 x 2withinsubjects: 5 per group, 5 in total (samepeople in each group) - 2 x 2, mixed design (betweenandwithin subjects):5 per group x 2 groups, so 10 in total Between subject designs require the highest numbers of subjects: WEEK 11 CORRELATIONAL RESEARCH DESIGNS Two broad disciplines of scientific psychology (according to Cronbach): 1. C orrelational:Concerned with studyingindividualdifferencesand investigating the relationshipbetweennaturally-occurring variables 2. Experimental: Not usually interested in individualdifferences, but inminimising these differences.Looks atgroupsandsimilaritieswithin them. Variables are assigned by the experimenter. Correlation and regression - orrelationidentifies anassociationbetweentwonaturally-occurring variables C (co-relation) - There is a positive association between number of hours studied per week, and marks at the end of the year - Regression:used tomake predictionswhenstrong correlationsexist - Y=a+bX a = y intercept, b = slope y = the dependent variable, x = independent variable e.g. Marks = error + b(hours studied) Need to know conditions for causality (correlation doesn’t equal causality): - covariation: they vary together - temporal precedence: one comes before the other - elimination of possible 3rdvariables Two specific problems: 1. W hichcomes first?(Directionality/temporal precedence)Does watching violent TV lead to aggression in children? Or do already aggressive children prefer to watch violent TV? 2. Third variables Violent parents may expose their children to more violent media ne solution to the directionality problem: cross-lagging O Eron, Huesman, Lefkowitz, & Walder (1972): Does watching violent TV cause aggression? - easured preference for watching violent TV programmes M - Measured peer ratings of aggression - 875 third grade students - Returned 10 years later, measured the same variables Caution: design can’t be inferred from the statistics Comparing introverts and extraverts on their level of obedience to authority: 1. A dminister tests of introversion/extraversion and obedience of authority; correlate the two 2. A dminister tests of introversion/extraversion; select 25% most introverted and extraverted; put them in an obedience situation and measure obedience; use a t-test or 1-way ANOVA. So why do we use correlational research? Practicality: - ome variablescan’t be randomly assigned(gender,age, personality variables) S Some research is conducted withpredictionin mind - E.g., predicting why certain people will do well on the job Ethical grounds: - Can’t randomly assign to brain damage or to experiencing childhood abuse etc Places where correlational research is used 1. Psychometrics (test development) - Testing reliability: split-half, test-retest.Mustbe highly correlated to indicate that they halves measure the same construct. - Testing validity: criterion validity.The abilityof the measure to predict an outcome. 2. Research in Personality and AbnormalPsychology (almostall is correlational:can’t assign people to having mental health problems or certain personalities) 3. Studying the nature-nurture controversy (correlation between particular genes/environment and outcomes) 4. Any cross-sectional study (e.g. correlation between exercise and stress levels) – can’t infer causality In summary: Correlational research: - ontributes a great deal to psychology, oftenwhenexperimental procedures cannot C be used - With modern, sophisticated statistical procedures, more complex questions about cause and effect can be addressed than in the past - Much correlational research takes placeoutside thelaboratory– for instance, quasi- experimental research and programme evaluation QUASI-EXPERIMENTAL DESIGNS An experiment is… - here participants arerandomly assigned W - Tomore than onecondition. AKA a randomised controlledtrial. efinition of quasi-experimental designs: D Quasi-experimental designs are those in which participantscannot be randomly assigned to conditions Or quasi-experiment exists whenevercausal conclusionscannot be drawnbecause there A isless than complete controlover the variables inthe study on’t have random assignments and theconditions aren’tequivalent, but they resemble D experiments in other respects (have anIV). Participantsare assigned to groups according to a characteristic they already possess (age, gender, IQ) which is the IV. eed to trymanually equatingthe groups oreliminatingother differences as N explanationsfor observed effects. Advantages: - Increasedexternal validity - Experiments are held under carefully controlled,artificial environments that may not generalise outside of these settings (i.e. in the real world) - Quasi-experiments are held undernaturally occurringconditions - Very useful forpolicydecisions / evaluation Disadvantages: s the quasi-independent variable isn’t completely under control and there’s random A assignment or control group, the research is muchmore vulnerable to threats of internal validity, specifically: - istory H - Maturation - Selection bias - Regression to the mean Two types of quasi-experimental designs to focus on: . Non-equivalentcontrol group designs 1 2. Interrupted time seriesdesigns T = treatment / intervention 1. Non-equivalent control group designs - urpose: evaluateeffectivenessof atreatment programmeby including a time P series component along with a “control group” that’s not exposed to the intervention - Sample size = 2 - The 2 groupsdiffer, they’re not randomly assigned.Validity therefore compromised. Therefore it is important to try to matchthe groupsas closely as possible prior to the study. Also vulnerable to regression to the mean and maturation. - Measure the two groupsbefore and afterthe intervention 2. Interrupted time-series design - urpose: to observe behaviour over timeprior to andimmediately afterintroducing P an intervention - Advantage: allows for theevaluation of trends - Often used inpolicy or media evaluation - E.g.: measuring the effect of an anti-smoking campaign - Need a lot of data points to actually understand the effect of the intervention. Important to measure the DV before as well as after the treatment: Example: effect of an incentive plan on worker productivity - racked productivity for 4 years before the intervention T - Tracked productivity for a further 6 years after ystematically evaluated possible threats to internal validity to verify the effectiveness of the S treatment Small N designs ery similar to time-series designs. Both useverysmall sample sizes.Used often in early V days of psychology (e.g. studying oneself) - mall N designs can be both experimental and quasi-experimental, depending on the S level of control over the independent variable. Some designs in this category involve true experiments (experimental), while others may have less control over the manipulation (quasi-experimental). Advantages of using small n designs: S ometimes used becausepotential subjects are hardto find(e.g. studyingrare disorders) Groupeddata can lead tomisleading results(experimentsaverage outresults). Small n designs are more interested inindividual differences - Averaged-out results from group designs may notaccuratelyportraythe response ofanyparticular participant (e.g. USA womenhave an average of 2.09 children) - Group means may have no counterpart in the behaviour of individual participants - Group means may obscure the fact the IVaffected someparticipants but not others ften more than one individual is studied (usually 3-8), but theirresults are analysed O separately and rarely averaged. ingle-participant researchers studyintraparticipantvariance: variability in an individual’s S behaviour within the same situation on different occasion Small N designs in applied behavioural research closely wedded to the study ofoperant conditioning - skinner’s influential research operant conditioning involves single participant designs lso been used to study the effects of various schedules of reinforcement and punishment a on behaviour in applied research single participant designs have been most used to study defects of behaviour modification - techniques for changing problem behaviours that are based on the principles of operant conditioning single participants research has also been used in industrial settings and schools lso used for demonstrational purposes simply to show that a particular behavioural affect A can be obtained howing that thebehaviour of a single individualchangesas a result of thetreatment, S punishment or reinforcement. This requires: 1. T he target behaviour must beoperationally definedin terms ofeasily recordable events(usually frequency) 2. E stablish abaseline level of respondingagainst which the effects of the programme can be assessed (there’s a period of no intervention) 3. Introduce the treatment and continue monitoring Some types of small N designs . W 1 ithdrawal/reversal designs 2. Multiple baselines designs 1. Withdrawal/reversal designs - he simplest isthe A-B design(A = baseline / beforethe intervention, B = during the T intervention) - Can’t be too hasty in deciding that the intervention caused the change in behaviour: there could have been a 3rdvariable. Therefore, theinterventionshould be withdrawn to see whether the behaviour changes again. - Other threats to internal validity includematuration,testing, and history. - Can also have more complex withdrawal/reversal designs: A-B-A or A-B-A-B – allows for more robust conclusions 2. Multiple baseline designs - sed when a withdrawal design isnot feasible(whenthe intervention/treatment U changes something in the participants that can’t be reversedor whenmedical reasons forbid the withdrawalof a treatment) - Multiple baselines approach uses avarying time schedulethat allows the researcher to determine if the application of treatment is truly influencing the change in behaviour. - We mightvary the length of timein the initial baselinedetermination and then apply the treatment to determine if thechange in behaviourcorresponds with the introduction of treatment. ultiple baselines across subjects: Thestart of treatmentconditions is staggered M (started at different times) across individuals. Because treatment is started at different times we can conclude that changes are due to the treatment rather than to a chance factor. Evaluating small N designs W hat aboutexternal validity? Small N designs don’ttend to have very good external validity due to their small sample sizes. Generalizability islimitedto the representivity of the sample. does not capture the nuances of real-world situations where multiple factors may interact What aboutstatistical significance? Can’t alwaysbe sure of it What about interactions? Difficult to explore because interventions are usually introduced one at a time. What about studies that don’t rely on frequency of response? Ethical issuesregarding ABA designs: is it ethicalto withdraw apotentially helpful treatmentfrom a troubled patient to assure the researcherthat it is in fact effective? (e.g. anti- depressants for suicidal patients he causal inferences one can draw from single participant designs are often weak, and the T effects are of questionable generalizability, but such studies can provideindirect, anecdotal evidencethatparticular effectscanbe obtained. Surveys Two important aspects: - ampling S - Structuring the questions he sample is not drawn directly from the population, but fromsampling frames(i.e. the T form in which the sample becomes available to the researcher) Probability sampling Random selection:each person has anequal chanceof being chosento be in the sample - sed whenever you want to learn something specific about an identifiable group of U individuals (want to be able togeneralize back tothe population) - We need a representative sample in order to generalise about the population, otherwise the sample is biased - For a sampling method to be considered probability sampling, it mustutilize some form of random selection. In other words, researchersmust set up some process or procedure that ensures, with confidence, that the different units in their sample population haveequal probabilities of being chosen. . S 1 imple random sampling: - Like “putting names in a hat” - Each element has exactly the same chance of being selected, and the selection of each element was independent of the selection of a previous one. - reates samples that are highly representative of the population C Effective and practical, except: - When you want to look at systematic features of the population (e.g. when you want to do research into gender differences but there are more women than men in a class) - When the population isvery large 2. Stratified sampling: - sed when populations consist of non-overlapping sub-groups/strata e.g. age, socio- U economic status - The researcher divides the entire population into different subgroups or strata (if it isn’t already divided), thenrandomly selects thefinal subjects from the different stratausing some form of probability sampling. - Proportionate stratified sampling: proportions of important subgroups in the population arerepresented preciselyin the sample.The same proportion of individuals that can be found in the population are also in the sample. 3. Cluster sampling: - sed when studying large populations and a sampling frame isn’t available U - The population is divided into more manageable groups of elements called clusters - A “cluster” is asampling unit(aclass, a schooletc)within whichone samples randomly on-probability sampling N Any kind of sampling where selection of elementisn’tdetermined by the statistical principle of randomness. an be usedwhen generalisability is not the goal,but rather studying therelationship C between two variables is (e.g. stress and memory). Useful for testing theories about processes considered universal. 1. C onvenience sampling:Requesting volunteersfrom agroupof available people whomeet the general requirementsof the study(e.g. UCT’s SRPP program). Only need to be available and willing to participate. 2. P urposive sampling:Recruitingaspecific typeofperson (can overlap with convenience sampling). E.g. women who are economic migrants. Sampling isdirected towards elements of interest. Purposive sampling depends not only on availability and willingness to participate, but that the cases are typical of the population group being researched 3. S nowball sampling:When you start off with purposive sampling, then ask the participants to recruit anyone they know who matches the inclusion criteria of the study. Gradually accumulating a large enough samplethroughcontacts and references Survey methods Interviews - Advantages: - Comprehensive, yieldslotsof information - Usuallydon’t get problems with unclear informationormissingdata - Disadvantages: - Peoplerefuseto be interviewed, can’t befound, interviewerrefuses to go there - Cost, logistics - Interviewer bias(e.g., cross-race bias) Written surveys - se either open or closed questions U - Often use Likert-like scales - Can be sent through the post or online or administered in a group setting Phone surveys - serandom digit dialling(to avoid missing unlistednumbers) U - Does excludethat portion of the population thatdoesn’t have a phone - Non-response ratecan be high NB: Evaluating survey research . Was the sampling framebiassed? Did itleave outcertain groups?. Was there asocial desirability bias? Participantsworry about how they look in the eyes of the researcher. E.g. parents unlikely to admit they do (or how often they do) spank their child. . Problems with questions: - mbiguity. E.g. “Visiting relatives can be fun.” A - 2-partquestions. E.g. “Is it wrong for female patronsin bars to swear and to buy drinks for men they don’t know?” - Framingof the question. E.g. “Do you prefer yourhamburgers flame-broiled or fried?” VS “Do you prefer a hamburger that is grilled on a hot stainless steel grill or cooked by passing the raw meat through an open gas flame?” Observational Research Types of observational research: . C 1 ase studies 2. Naturalistic observation/Participant observation 3. Archival research . C 1 ase studies - A detailed, descriptive study of a single individual, group or event (i.e. case) - There’s a tension in psychology between theidiographicand nomothetic approaches - Idiographic = individual. - Nomothetic = the group. - Case studies studying the individual are most common, but researchers can sometimes perform case studies of groups - The data can come from a variety of sources such as observation, interviews, news reports, and archival records Value of the case study: - ourceof ideas andhypotheses(e.g. Freud and hisclients, Piaget and his own S children theories) - Source fordeveloping therapy techniques - Permit the study ofrare phenomena(hard to findenough people for a group) - Can provide acounterfactualfor notionsconsidereduniversally applicable(e.g. in criminal profiling) - They havepersuasive and motivationalvalue (in theadvertising industry) Limitations of the case study - hey’re virtually useless in providing evidence to test theories, as the case study T occurred in anuncontrolled environment. No matterhow plausible the explanation offered, alternative explanations cannot be ruled out (internal validity). - The heavy reliance onanecdotal informationmakesfor the possibility of quite biassedpresentation. There’s no relevant comparisoninformation and control over other variables. - Lack ofgeneralisabilitytootherindividuals orsituations. - Observer bias: in behavioural science, the researcheris often the participant’s psychotherapist. There’s therefore no way of determining thereliability or validity of the researcher’s observations or interpretations.Additionally, the researcher often has a stakein the outcome of the investigationof self-fulfilling prophecies. 2. Naturalistic + participant observation - he attempt tostudy normal behavioursof people or animals as they act in their T everyday environments. - The chief method of anthropologists and wildlife researchers - The researcher… - Could behidden(e.g., behind a one-way mirror; ina bird-hide) - Could just sit on a bench in a strategic location - Could just analyse avideo - Participant observation:join the groupbeing observed .g. The theory of cognitive dissonance emerged out of participant observation. Researcher E “joined” a cult and wrote a book about it. Issues with this study: - Joined under false pretences - Gave the other cult members the feeling that their ideas may be correct, because 7 other people (the researchers) had joined them valuating observational methods E Absence of control: - Conclusions aboutcausality cannot be drawn an serve the purpose offalsification(show thatwhat are considered universal laws aren’t C actually) Observer bias: - Allowingpreconceived ideastocolour one’s interpretationsof one’s observations Subject reactivity - Subjects mayreact to the researcherbeing present,taking notes etc xample: observations of touching E . Hall & Veccia (1990): Men and women in public places: who initiates the touching? Are there gender differences in the type of touching? Are there age differences? . Mixed-sex and same-sex touching in a variety of public places (airports, subways, malls) in Boston . Handheld tape recorder and a timer that beeped every 10s . A 3 rchival research - Archival data range frompublic informationsuch ascensusdata, court records, genealogical data, etc. through moreprivate informationsuch as credit histories, medicalrecords,educationalrecords, and diaries. Evaluating archival research Advantages: - nlimitedinformation available; the only limitation is the researcher’s creativity U - No subject reactivity(as you’re not working withactual participants) - Canconverge with lab research, thereby increasingexternal validity (testing whether lab- generated ideas correspond with archival data i.e. the real world) Disadvantages: - ecause it already exists, some piecemay be missingor may not be B representative - Experimenter bias: selecting only those records thatsupport one’s hypothesis Ethics in Psychological Research The start of formalising ethics - The Tuskegee syphilis study - Black men with syphilis living in a poor area were studied. They were told they were being treated (when they weren’t) and the study was continued even after effective treatment was developed - Study started in the 1930’s but was only made public in 1968 ed to: L TheBelmont Report:Ethical Principles and Guidelinesfor the Protection of Human Subjects of Research(1979) NB: Three fundamental principles: 1. B eneficence:maximise benefitsandminimise risks(benefits should outweigh the risks) 2. Autonomy:people are treated ascapable of makingdecisionsabout whether or not toparticipatein research(adults give informedconsent, children give assent). Confidentiality is also important. 3. Justice:fairness(fairness about carrying the risks/burdensof the research; fairness about receiving the benefits). People shouldreceivewhat is due to them. Participants should be treated withfairness at allstages of the research(includes fair selectionand providingcare/supportdistress/harmedparticipants). Need to be able to define and explain the above. eception may only be used when the study is not harmful, there are no other alternatives, D and the research is important. Participants must be debriefed afterwards. The IRB and Ethics codes - Institutional Review Board – generic name Draws on: - heAmerican Psychological AssociationEthics Code (see T www.apa.org/ethics) - TheHelsinki Declarationof the World Medical Association B: Informed consent N Informed consent deals mostly with autonomy. Participants need to understand what they’re signing up for. Knowledge - nderstanding thenatureof the experiment (time requiredetc.), thealternatives U available, and thepotential risksandbenefits Volition - articipants must provide their consentfree fromconstraint or duress(can’t be P forced), and mayrevoketheir consent at any time(NB when working with children, prisoners etc.) - “Thank you” gifts or potential benefits shouldn’t be so big that they make participants feel that they can’t drop out. Participation needs to be entirelyvoluntary. Competence - he individual’sabilityto make awell-reasoned decisionand to giveconsent T meaningfully - People in comas, people with intellectual disabilities Consent needs to be formalized in writing unless the risks of harm are very low. Consent forms - verview O - Description ofprocedures - Risksand inconveniences - Benefits - Costs andeconomic considerations - Confidentiality - Alternative treatments - Voluntary participation - Questions and further information - Signaturelines Other ethical issues in research - ata fraud D - Allocation of credit i.e. plagiarism. Stealing someone else’s work/ideas - Sharing of materials and data Basic single case experimental designs three basic single case experimental designs: ABA, multiple-I, and multiple baselines designs . A 1 BA designs - most common - attempts to demonstrate that an independent variable affects behaviour by first showing that the variable causes a target behaviour to occur and then showing that the removal of the bearable courses behaviour to cease - these are sometimes called reversal designs participant is first observed in the absence of the independent variable Target behaviour is measured many times during this phase- to establish a baseline for comparison T he independent variable is introduced and the behaviour is observed again if the independent variable influences behaviour you will see a change in behaviour from the baseline to the treatment period - however despite this the researcher should not be too hasty to conclude that the effect was caused by the independent variable - some other event occuring at the same time as the independent variable could’ve produced the observed effect to reduce this possibility the independent variable is then withdrawn if the independent variable is in fact maintaining the behaviour of the behaviour may return to its baseline level . m 2 ultiple- Interval designs - Single case experimental designs that present varying nonzero levels of the independent variable or cold multiple I designs researches insert a baseline period between each successive an introduction of a level of the independent variable resulting in a ABACA design - commonly used in research that investigates the affects of drugs on behaviour - also used for graduated exposure . m 3 ultiple baseline designs - two or more behaviours are studied simultaneously after obtaining baseline data on all behaviour is an independent variable is introduced that is hypothesised to affectonly oneof the behaviours - by measuring several behaviours the researcher can show that the independent variable cause the target behaviour to change but did not affect the other behaviours - if the effect of the independent variable can be shown to be specific to certain behaviours the researcher has increased confidence that the obtained effects were in fact due to the independent variable
