Revision Deck Flashcards

Question

What is differential attrition?

Answer 1

When the rate of participant loss over conditions differs systematically and significantly across groups.

Answer 2

When the pre-test causes an effect rather than the manipulation/intervention

Answer 3

a)Vary the treatment between-participants and randomly assign participants to levels of the treatment IV (treatment vs. no treatment). With random assignment to groups, a pre-test is not necessary for control  can avoid the problem. (b) Use a Solomon Four-Group Design. Half get pre-test, half don't

Answer 4

occurs when exposure to the IV or DV (or associated procedures) causes changes in responding that provide an alternative explanation for the results.

Answer 5

Experiments on attention & performance, cognitive processes. Ps complete multiple questionnaires on attitudes, beliefs, etc.

Answer 6

(a) Randomly sequence conditions. (b) Counterbalance condition orders (c) Vary IV between-participants. (d) Used fixed sequence that works against the hypothesis. (e) Use a yoked control.

Answer 7

Counterbalancing ensures that: each level of the IV occurs equally often in each available slot, and every condition precedes every other condition equally often.

Answer 8

A participant’s response to a stimulus item may be influenced by the extent of the contrast between the present item and the previously-presented item. A large change in the IV value from one item to the next may lead to a disproportionately large shift in responding. Counterbalancing will not account for contrast effects as contrast is not equal

Answer 9

if A2 is affected by A1 more than A1 is affected by A2 | An issue as counterbalancing will not equally alter carryover effects

Answer 10

Time- and place-dependent changes that: happen to occur during the course of the study, and provide an alternative explanation for the results. History confounds can affect: within-participants IVs (mainly). between-participants IVs.

Answer 11

Time-dependent changes that: occur within participants during the course of the study, and provide an alternative explanation for the results.

Answer 12

Control group that receives placebo or no treatment

Answer 13

Instrumentation change is: a shift in measurement over time in the direction predicted by hypothesis, resulting in apparent change in the DV. Preventable by random assignment, concurrent score collection and comparison group

Answer 14

An operations confound is an alternative explanation (specifically, a competing construct) that is introduced in the procedures used by the researcher to operationalise a construct.

Answer 15

A materials confound is an alternative explanation (specifically, a competing construct) that is introduced in the materials used by the researcher to operationalise a construct.

Answer 16

Replicate over stimuli. Random assignment of items to levels of the IV Hold extraneous item variables

Answer 17

when spurious results are produced due to: inappropriate analysis of data; or inappropriate transformation of scores (i.e., preparing data for analysis).

Answer 18

Simpson’s Paradox Can occur when two tests differ in both: (a) difficulty; and (b) the proportion of Ps undertaking each test. e.g., if the less able group of participants is over-represented in the easier test, it can look as though they are performing better overall if the scores are inappropriately proportionalised Drawing conclusions about % passing is misleading if raw frequencies are pooled over the tests

Answer 19

Observer expectancy issues occur when an observer is biased to perceive participants’ behaviour as consistent with the hypotheses. Where possible, use observers who are blind to the hypotheses and conditions (i.e., which participant is in which condition). Videotape participants’ behaviour for independent observers to code. Where possible, use objective/unambiguous measures of behaviour (e.g., a physiological index).

Answer 20

Experimenter expectancy issues occur when the experimenter communicates to the participant what to do to support the hypotheses. Experimenters may do this: through unintentional nuances in: tone of voice, expression, wording of instructions, etc..; through the “demand characteristics” of the settings, stimuli, tasks, questions, etc., used in the study

Answer 21

Where possible, use experimenters who are blind to the hypotheses and conditions. Include procedures that disguise the purpose of the study

Answer 22

Participant expectancy effects may take several forms: Demand Compliance (the participant responds in accordance with demand characteristics). Participant Resistance (the participant tries to resist what they think the experimenter wants them to do). The Hawthorne Effect (When the treatment group changes their behaviour not because of the treatment itself, but because they know they are getting special treatment). Positive Self-Presentation (participants respond in a way that portrays them in a good light)

Answer 23

Prevent experimenter expectancy/demand characteristics Use a placebo-control Blind participants to the hypotheses and conditions.

Answer 24

Floor Compression: occurs when a test is too difficult or inadequate time is given – scores are compressed near the minimum. Ceiling Compression: occurs when a test is too easy – scores are compressed near the maximum. usually works against a significant effect (i.e., a sensitivity issue). is a confound if null results were expected.

Answer 25

Correlated DVs are sometimes combined in a weighted linear combination or (unweighted) average, to produce a composite dependent variable. Correlated DV measures do not necessarily measure the same construct (i.e., “the same thing”). Consequently, individual components of a composite dependent variable may behave differently over IV levels.

Answer 26

systematic variation (potential confound). vs. variable held constant or restricted in range (potential generality limitation).

Answer 27

it is held constant (or restricted in range) in the study, | and, if varied systematically, it would interact with IV(s) in the study

Answer 28

(a) The Obtrusiveness of Scientific Trappings. (b) The Physical Setting. (c) Researcher Attributes. (d) Researcher Expectancies.

Answer 29

(a) The Artificiality of Laboratory Experiments. (Variables held constant in the lab may exert large effects in the field. Conversely, variables manipulated in the lab may have little practical influence in the field.) (b) Pre-Testing Effects. (c) Carryover Effects. (d) Demand Characteristics.

Answer 30

(a) Convenience Sampling. (For generality, a representative sample is required) (b) Psychology Student Samples. (c) Volunteers. (Their results may not generalise, because volunteers may exceed the norm in) (d) Attrition.

Answer 31

(a) Changes in Social Norms, etc.. (b) Changes Within Psychology. (Past results may not be replicated by experimenters with different assumptions and approaches.) (c) Type I Errors & Artifacts; Short-Term Circumstances. (May produce results that will not be replicated)

Answer 32

Replicate lab findings in the field, use a variety of designs and approaches To establish/improve spatial generality, conceptual replication is the best approach. Conceptual replication = replication with variation For temporal generality, replication is essential

Answer 33

``` systematic variation (potential confound). vs. random variation (may affect sensitivity). ```

Answer 34

Null results are ambiguous. Possibilities: The hypothesis is incorrect. The study was not sufficiently sensitive.

Answer 35

Ensure adequate variation of the IV. Methods include: (a) Use extreme levels of the IV. (b) Use parametric variation. (Stepping systematically through the estimated effective range of the IV.) (c) Check the potency of the manipulation. Tailor the procedures for maximal IV effect. (Optimise the level of a non-focal interactive variable (i.e., a factor which is not an IV in the study, but which has the potential to interact with one of the IVs)

Answer 36

Adequacy/Sensitivity of a Measure = The extent to which the measure is capable of tapping into (or registering) variation in behaviour.

Answer 37

(a) Check score distribution, range, skew, etc.. Check whether the range of scores is adequate, etc.. Ideally, do so before the study (i.e., in a pilot study). (b) Conduct a power analysis. Assesses ability to show a significant difference (e.g., between groups) if such a difference truly exists. Can use to decide how many participants to recruit. (c) Pilot test for possible floor/ceiling effects. Caused by insensitive tests, e.g., ability test: Too difficult (floor effect) or too easy (ceiling effect). (d) Make measurement more fine-grained. (e) Consider whether participants are likely to use the entire range. (f) Conduct a “Benchmark” Check. i. e., Replicate an established phenomenon, using new IV tasks, procedures, etc., to demonstrate their adequacy.

Answer 38

Construct Validity and Eliminate any Alternative Explanations

Answer 39

Do the IV(s) and DV(s) capture the construct(s) they are intended to tap into?

Answer 40

Face Validity. Criterion Validity. Convergent Validity. Discriminant Validity

Answer 41

The degree to which scores on the measure are correlated with scores on a measure of a theoretically-related construct.

Answer 42

Whether the measure correlates well with other measures of the same construct.

Answer 43

When the same measurement method is used for two measures, part of the covariation between them may be attributable to the common method itself.

Answer 44

Whether a measure is: uncorrelated with measures of unrelated constructs, or separable/distinguishable from (i.e., not highly correlated with) measures of related or similar constructs. Discriminant validity is important; convergent validity alone is insufficient to establish construct validity.

Answer 45

``` Face Validity. Parametric IV Variation. Parallel DVs. Manipulation Check. Benchmark Check. ```

Answer 46

As for sensitivity, Varying an IV Parametrically = Stepping systematically through the estimated effective range of the IV. But, for construct validity, Q is whether the effect varies as expected if the IV has captured the intended construct.

Answer 47

Replicate an established phenomenon, using new IV tasks, procedures, etc., to demonstrate their adequacy. Does the IV task produce recognised effects? Is the DV responsive to variables previously established as influential?

Answer 48

The confound can be varied separately from the IV; or | It is possible to make the influence of the confound comparable across conditions.

Answer 49

Add conditions, IVs, DVs, measures, etc., to a study (or a series of studies) to produce different patterns of results for competing explanations

Answer 50

It involves the same strategy used for validating constructs, but with an emphasis on eliminating a particular variable as an explanation.

Answer 51

Variation and Replication basically involves investigating the same causal relationship between constructs using multiple empirical approaches. ``` Every approach (every study) has weaknesses, so exact replication is not useful for eliminating alternative explanations Different methods also have different strengths. ```

Answer 52

There are several approaches that all involve augmenting the design to generate opposing predictions to the effect that: If the hypothesis is correct, pattern A will be obtained, or If a nuisance confound or rival hypothesis is responsible for the phenomenon, pattern B will be observed.

Answer 53

Add a condition that will produce an effect according to the alternative interpretation, but not according to the hypothesis.

Answer 54

Add an IV that will interact with an existing IV according to the theory but will only show a main effect according to the alternative explanation.

Answer 55

Add DV(s) which will be affected differently by the IV(s) according to alternative explanations.

Answer 56

(a) Check Covariation of Confound with IV or DV. (b) Assess Participants’ Awareness of Aims. (c) Check for Observer Agreement & Validity.

Answer 57

Check Covariation of Confound with IV or DV. If the suspected confound is not correlated with both the IV and the DV, it cannot confound the results. Assess Participants’ Awareness of Aims. Use a post-experimental questionnaire to check whether the participants worked out the hypothesis (e.g., Petty et al.). Check for Observer Agreement & Validity- obvious

Answer 58

Data correction techniques involve adjusting the raw data before conducting the focal analyses, to remove the influence of a potential confound.

Answer 59

Removing the affected data, like lost participants to attrition, participants who figured out the hypothesis and data affected by regression simply removes these confounds Avoids data linkage issues.

Answer 60

Two strategies: Express Raw Frequencies as a proportion of opportunities ``` Comparing frequencies (e.g., the number of participants who pass a test) between groups of participants: Raw frequencies can be misleading if there was a different number of participants in each group. ``` And Expressing Individual Ps' Scores as a proportion of opportunities In some research studies, the number of opportunities an individual participant has to contribute to their own DV score can vary from participant-to-participant. There is a potential confound (or counter-confound) if the average number of opportunities per participant varies systematically from one condition to another. Hence, individual DV scores are sometimes expressed as a proportion of some other performance measure (i.e., to quantify the number of opportunities that the participant had to contribute to their own DV score).

Answer 61

ANCOVA (Analysis of Covariance). Partial Correlation Techniques. Statistical Modelling Techniques

Answer 62

(a) Under-Correction. The effect of the covariate may not be fully corrected. Because of the unreliability of measures, the confound may still be contributing to group differences on the DV (b) Assumptions of ANCOVA. The assumptions of the test must be met, or else the results may be spurious. Most importantly, the regression slopes (predicting the DV from the covariate) must be the same for each group (strong assumption). (c) Additional Confounds. ANCOVA only controls for the covariate. Another confound of the IV-DV relationship may remain. Therefore, ANCOVA cannot compensate entirely for non-random assignment. (d) Efficiency Issues.

Answer 63

Partial Correlation Techniques can be used to remove the effect of the confound from the IV and/or DV when assessing the IV-DV correlation. Involves essentially the same ‘residualising’ technique used in ANCOVA

Answer 64

Statistical Modelling Techniques can be used to: Control for the effects of confound(s). Assess the relative importance of theoretically relevant competing constructs. Examples: (a) Multiple Regression. (b) Path Analysis. (c) Structural Equation Modelling

Answer 65

``` Path analysis is a regression-based technique used to assess mediated and direct effects among a set of variables. The observed pattern of correlations is compared against the pattern predicted by a causal model (based on theory). Assess sign (+ vs. –ve) & significance of path weights. Test for ‘goodness-of-fit’ of data to model. ```

Answer 66

Structural Equation Modelling (SEM) is similar to Path Analysis, but more sophisticated. Main difference: Paths are drawn between latent variables rather than observed variables (i.e., IVs & DVs). Latent variables capture the common variance among multiple measures of a construct.

Answer 67

(a) Factors Affecting Strength of Relationships i. Unreliability of Measures. Stronger relationships may indicate nothing more than higher reliability. ii. Common Method Variance. e.g., If all measures are self-report. May inflate some associations. iii. Chance Results. Statistical modelling is not useful with small samples. iv. Our Ability to Think of Possible Models. Need theoretically-motivated models – not post-hoc! (b) Making Causal Inferences Although Path Analysis/SEM may be referred to as ‘causal modelling’, they do not demonstrate causation. Causation cannot be established by statistics alone

Revision Deck Flashcards

(91 cards)