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;
35 Cards in this Set
- Front
- Back
Elevated F Score:
|
65+ = random responses, deliberate malingering, eccentric or all t, all f
(tells validity) |
|
Low/High L Score:
|
Low: frank or insightful, or exaggerate badness
High: Denial, lying, or lack of insight tells you if valid. |
|
K Scale: (correction) High/Low
|
High K: defensive or fake good, associated with poor prognosis
Low K: frank, self-critical, fake bad. "Suppressor variable" corresponds with education, SES, defensiveness |
|
? scale: High
|
high ? scale suggests readng problems, indecisive, distractible, rebellious, defensive.
|
|
True Experimental vs. Quasi Experimental:
|
Experimental you are controlling assignment of subjects to tx groups, QUASI-EXPERIMENTAL you are not. (ie, cluster sampling?)
|
|
Random Assignment vs. Random Selection
|
Random assignment: allows you to be sure IV caused DV, random selection allows you to generalize this outward. Random assignment is true experimental.
|
|
5 Ways to control extraneous (confounding) variables:
|
Confoundign variable: irrelevant to purpose of study, but has an effect (ie, level of depression)
1. Random assignment to tx group 2. Hold it constant- (ie, use all mildly depressed patients) 3. Match subjects on the variable (ie, one mild in each group, one moderate in each group) 4. "Blocking" build it in- lump and match, rather than individual match 5. Statistical Control (ANCOVA) |
|
ANCOVA:
|
Allows you to remove variability of DV due to the confounding variable.
Equalizes all subjects on the variable, good in quasi-experimental. Comparing a bunch of samples, want to get out confound (ie, looking at school lunches on health, but confound is how much exercise) (an-CONFOUND-ova) |
|
Interrater Reliability
|
Kappa Statistic
|
|
Split-Half Reliability
|
Use: Spearman-Brown Prophecy (once youve done part)
|
|
Internal Consistency Reliability:
|
Use Coefficient alpha, or if dichotomous, use Kuder-Richardson 20 (KR20)
(internal consistency reliability not good on speed tests: speed is the opposite of power) |
|
Stability
|
AKA Test-Retest Reliability
|
|
Standard Error of Estimate:
|
Think Regression Line
|
|
Standard Error of Measurement:
|
To construct Confidence Interval around a score.
|
|
Convergent & Discriminant Validity
|
Convergent: Does it match other
Forms of construct validity ideas of measurement- ie, other constructs of "depression", ie, is it not confusing depression with anxiety? Divergent: Does it NOT match other things that are NOT depression USE: MTMM Matrix |
|
MTMH Matrices
|
For Convergent/discriminant
|
|
Monotrait Monomethod
|
Reliability Coefficients
|
|
Monotrait-Heteromethod
|
Correlation between different measures of same trait: (ie, the bdi and the phq-9) so high = convergent
|
|
Hetertrait-Monomethod
|
Correlation between different traits and the measure- ie, if low it has discriminant validity
|
|
Heterotrait-Heteromethod
|
correlation between different traits that have been measured by differed methods- also show discriminant validity if small.
|
|
Factor Analysis: Why?
|
Find minimum number of common factors to account for intercorrelations in set of tests.
To find construct validity, convergent and discriminant validity. |
|
Factor Analysis: How
|
1. Administer a bunch of tests
2. Correlate scores on each test with every other one to create a matrix ("R") 3. Using magic, convert correlation matrix to factor matrix 4. Simplify the factors by "rotating" them" 5. Interpret and name the factors |
|
Factor loadings
|
The correlation coefficients on a factor matrix - ie, the correlation coefficient of test A and the factor. (ie, test A's measure of assertiveness)
|
|
Communality:
|
The "common variance" in factor matrix: ie, the column after the tests and the two factors that shows the common variability of the two factors. ie, 64% variance is explained by combo of factor I and II.
|
|
orthogonal rotation:
|
Resulting factors (in factor analysis matrix) are uncorrelated. Used if you think it WONT be related.
|
|
Oblique rotation
|
Resulting factors ARE correlated, and attributes are NOT independent. Used if you think it will be related.
|
|
Hetero vs. homoscedacity
|
How much regression line clustering starts to change shape- ie, does it start to widen/ turn into a "v" or tornado shape?
|
|
Reactivity / Reactive arrangements
|
When participants change behavior because they are being observed- figure it out due to demand characteristics.
|
|
Regression line: (correlation)
|
Where the clustering happens, and you get a correlation between -1 and 1. (then you take that number and square it to get %variability accounted for by the IV.)
|
|
Counterbalanced design:
|
different subjects get different IVs in different order to to get rid of order and carryover effects.
|
|
Factorial Design:
|
2+ IVs. Analyze each IV and their interaction. (ie, cbt, medication and their combined power, or mild, moderate and medication)
|
|
Aurocorrelation
|
A confound- screws you up because people will correlate pre and post test on measurements, and increasing type 1 errors.
|
|
Ordinal Scale:
|
Ranks, numbers, more/less, likert scale, but dont tell us HOW MUC difference between. (ie, how different is a 10 and 8 from a 2 and 4 on likert)
Central tendency is mode or median |
|
Interval:
|
Like ordinal, but intervals are set- ie, a WISC has same amount of difference between 80 and 85 as between 115 and 120. So can be added/subtracted.
Standardized scores. Central tendency is M, M, or M. |
|
Ratio Scale
|
Has an absolute 0: ie, 100$ is twice 50$. (reaction time, # correct on test, frequency of a behavior). Can be multiplied/divided.
|