Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
63 Cards in this Set
- Front
- Back
|
Validity = |
Truthfulness |
|
|
Construct Validity |
Do Independent and Dependent variable measure what they are supposed to? |
|
|
Reactivity of Subjects (participants) |
Subjects' expectations may bias behavior
|
|
|
Hawthorne Effect
|
Knowing you are being observed influences performance beyond any effect of the independent variable |
|
|
Solutions |
Deception: misled subjects about the true nature of the experiment "Blind" and "Double Blind" procedures |
|
|
Problems with validity |
Confounding variables Random error of measurement |
|
|
To minimize confounding variables |
Operational definitions Protocols- rules for observing behavior, handling subjects, other methodological details. |
|
|
External Validity |
Are results generalizeable across subjects, variables, setting? Subject Representativeness Variable Representativeness Setting Representativeness |
|
|
Subject Representativeness |
Are white rats and college students representative? Are underlying processes same as in other species or populations? e.g. learning in Skinner box vs. learning in school |
|
|
Variable Representativeness |
How does one select variables to study? Personal interest/curiosity Literature search Theory Is the variable we have chosen representative? Repeat the experiment with minor variations in the variable (e.g. different type of stressor) |
|
|
Setting Representativeness |
"Ecological validity" Nor concerned with "realism" Results should generalize to real world Repeat observations in a natural setting (i.e. field study) Be careful in generalizing Replicate the results under different conditions to establish external validity |
|
|
Internal Validity |
Validity of making casual or explanatory conclusions Undermined by confounding |
|
|
Reliable= |
Consistent |
|
|
Reliability of Measurements |
Every set of measures contains some variability (random error) Measured score= true score + error e.g., IQ score = True IQ + (Effects of fatigue, general health, guessing) RT = Latency inherent in nervous system + attentional/motor factors ↑Variability (error) = ↓ Reliability |
|
|
Statistical Reliability |
How likely are results due to chance? If less than 1 in 20 (p<.05) reject possibility of chance As the number of observations increases, so does the reliability |
|
|
Experimental Reliability |
Does a replication of the experiment yield the same results? |
|
|
Test Reliability |
Take same measures on different occasions "test-retest reliability" If practice effects, give "parallel forms" of test Can divide test items in half and correlate scores on the two "split-half method" |
|
|
Sampling |
Population vs. sample Is sample representative? |
|
|
Random sampling
|
Each member of the population has an equal chance of being selected for the sample Large sample better than small Is this feasible? |
|
|
Random Assignment |
Assign subjects to experimental conditions on a random basis Minimize confounding |
|
|
Measurement |
How we assign numbers to what we measure determines the conclusions we can draw |
|
|
Measurement scales |
Differ with respect to: Magnitude of attribute (e.g., relative vs. quantitative) Intervals between values (equal vs. unequal/unknown) Zero point (true zero point vs. arbitrary) |
|
|
Types of scales |
Nominal Ordinal Interval Ratio |
|
|
Nominal scale |
Sorts into categories Statistics (e.g. mean) have no meaning e.g., rating scale |
|
|
Ordinal scale |
e.g., ranking order Can assign magnitude, but intervals not equal |
|
|
Interval scale |
Magnitude can be assigned Equal intervals (35-40=85-90) No real 0 point |
|
|
Ratio scale |
Magnitude Equal intervals Real 0 point Ratios of values are meaningful (200lbs=2 x 100lbs) |
|
|
Experimental Design Two basic designs |
Between subjects-different groups get different treatments Within subjects-all subjects get all treatments |
|
|
Between Subjects Design flaws |
Avoid carry over effects Potential confound- a priori differences between groups |
|
|
Solutions to the the Between Subjects Design |
Matching Randomization |
|
|
Matching |
Subjects on some criterion, then random assignments to groups But may create mismatch on other criteria Subject attrition |
|
|
Randomization for between subjects |
Throw of the die Random number generator |
|
|
Within subjects Design Flaws |
Major drawback-carry over effects |
|
|
Solutions to the Within Subjects Design |
Randomization Counterbalancing |
|
|
Randomization for within subjects |
Randomize the order of treatments |
|
|
Counterbalancing |
Complete counterbalancing Each treatment occurs in each time period f the experiment e.g., 3 treatments (A,B,C) Problem: as number of treatments increases, number of orders increases disproportionately (formula=n!) Therefore can't run all orders |
|
|
Incomplete Counterbalancing
|
Each treatment occurs equally often in each portion of the experiment Balanced Latin Square Design |
|
|
Mixed Design |
A. One or more between groups independent variables plus one or more within groups independent variables B. Example: Effect of amphetamine and feeding condition on milk uptake Between groups: Bottle vs. Cannula Within Groups: Amphetamine vs. Saline |
|
|
Some Guidelines for choosing an Experimental Design |
A. If carry over effects are likely, use Between Groups design. (i.e., when carry-over effect is permanent or long-lasting; brain damage, toxic chemical, time-dependent variables, like training) B. If interested in changes in behavior over time, use Within Groups design (e.g., learning) C. If subject pool has markedly different individuals; or if you expect large individual differences in responses use Within Subjects design (e.g., screening new analgesics) |
|
|
Multifactor Experiments |
Advantage over single-factor experiments: they can yield interactions |
|
|
Between Groups |
Example: Pratkanis et. al., 1988-"Sleeper Effect" 2 X 2 factorial design 2 independent variable, each with 2 levels Factorial= all possible combinations of treatments are examined |
|
|
Between Groups Independent Variables: |
1. Delay between message and rating (a) 0 (b) 6 weeks 2. Cue Presentation (a) Before message (b) After message |
|
|
Within Groups |
All levels of each treatment experienced by all subjects Example 1: Dewing & Hetherington, 1974 - Solving anagrams 2 X 3 design |
|
|
Within Groups: Independent variables |
1. Imagery value of solution (a) High (b) Low 2. Hint (a) None (b) Structural (c) Semantic |
|
|
Problems of control (carry over effects) |
1. Solving one type of anagram with one kind of hint might affect solving another type of anagram with a different hint 2. Practice effects each subject had 2 trials |
|
|
Solutions to the Carry Over Effects |
1. Complete Randomization- use 12 random orders of anagram 2. Block Randomization- randomize order twice Each condition randomized for first 6 trials, than again for second 6 trials Advantage: every condition tested once before any one is repeated 3. Counterbalancing Generate a 6 X 6 balanced Latin Square for the 6 treatments Advantage (over randomization): each treatment precedes and follows every other treatment equally often |
|
|
Mixed Design |
One or more Between subjects independent variables plus one or more Within subjects independent variables Example: Effect of alcohol and the expectation of alcohol's effects on aggression Independent Variables: Alcohol dose (3 levels) 0, 0.4, & 0.8 g/kg Expectation (2 levels) Expect, Don't expect Dependent Variables: Number and intensity of shocks delivered Control Variables? |
|
|
Latin Square Alcohol dose chart |
|
|
|
If a study has external validity, one is entitled to |
Generalize |
|
|
A major threat to internal validity is |
Confounding |
|
|
Test reliability determined by a correlation between scores from two prts of a test is called |
Split-half |
|
|
The scale property that distinguishes an ordinal scale from a nominal scale is |
Magnitudes |
|
|
As compared to psychophysical scales, psychometric ones |
Are not as precisely defined on the input side |
|
|
A major disadvantage of between- subjects designs is taht |
One must use fewer independent variables |
|
|
Advantages of within-over between-subjects design include all of the following except that |
There is less chance of contamination between treatments |
|
|
In a completely counterbalanced experimental design |
All possible treatment orders are used |
|
|
In a balanced Latin square design |
Each treatment preceds and follows every other treatment equally often |
|
|
In an experiment designed to test the effects of alcohol on appetite, if drinks X and Y contain .5 and 1.0 ounces of vodka in orange juice, respectively, and drink Z contains only orange juice, then the control group in the study should receive |
Drink Z |
|
|
Each of the following is a reason for doing multi-factor experiments instead of single factor ones except |
Increased control over subject variables |
|
|
In a 2 x 3 factorial epxeriment using a between-subjects design, each subject serves in condition(s) out of ______ conditions in the experiment |
1;6 |
|
|
In the table below, the pattern of results indicates: |
Both main effect and the interation are important |
|
|
The major dancer in the use of complex within-subjects designs is: |
Carryover effects |
|
|
In a mixed design, there is (are) |
At least on withing-subjects independent variable and at least one between-subjects independent variable |
