Exam 2 Content Flashcards
Relationship of r with cov
Divide by product of standard deviations of X and Y to have r on a standardized scale from 0.00 to 1.00
formula for r
Σ(ZxZy)
N-1
How to interpret r?
Strength = 0 to 1.00
Direction = +/-
4 Characteristics of Scatter Plots
- Direction (+/-)
- Strength (0 to 1)
- Form (linear/curvilinear)
- Outliers/unusual features
Problems with r?
Restriction of range = doesn’t give full picture
Heterogeneous subsamples = doesn’t take into account different “weights” that may affect the slope (and r)
When to use rs?
Use when the data is ordinal
Formula for rs
1- (6ΣD2)
n(n<sup>2</sup>-1)
How to interpret rs?
Magnitude, direction, form, outliers & unusual features
Formula for line of regression
yhat=bx+a
b = r(Sy/Sx)
a=ybar+ bxbar
S = square root(Σ(x-xbar)/(n-1)
What is another name for IV?
predictor
What is another name for DV?
criterion
What does r2 give?
The amount of variance in Y (DV) accounted for in X (IV)
Formulae for r2?
r2 in simple regression
SSerror _ = Σ(Yi-Yhat)2_
SStotal = Σ(Yi-Ybar)2
SSregression =Σ(Yhat-Ybar)2
SStotalΣ(Yi-Ybar)2
Problems with linear regression?
Multicollinearity, heteroscedasticity, outliers, and extrapolation
What’s wrong with multicollinearity?
Variance is accounted for twice and not all of it is considered
What’s wrong with heteroscedasticity?
Assume a normal distribution, but it’s not…
What’s wrong with outliers?
They weigh the R and give an invalid depiction of the sample
What’s wrong with extrapolation?
Not accurate
What’s a moderator?
Moderators are 3rd variables that explain WHEN a relationship occurs (nominal variables, like gender, height)
How do you test for moderation?
Run regression for interaction term (IVxmoderator) and if sig is <.05, then there’s moderation (code moderator = 0)
What is a mediator?
3rd variable that explains WHY there’s a relationship between predictor and criterion
How to test for mediator?
Run regression between interaction term (IVxmoderator);
if standardized beta coefficient of c’ decreases, then there’s moderation
What is reliability?
Consistency in data and ability to reproduce the data
5 ways to test reliability?
- Test, Re-test reliability
- Alternate Forms reliability
- Split Half reliability
- Inter-rater Reliability
- Internal Consistency
What is test, retest reliability?
Test, then test again later
May induce practice effects :(
What is alternate forms reliability?
Test and test again with different & equivalent test
What is split-half reliability?
Run regression between 2 halves of tests (odds and evens, first and last part)
What is inter-rater reliability?
Regression between judges and their scores (uses intraclass consistency)
subj. variability/subj. variability + error
0 = no reliability
1 = reliability
What is internal consistency?
Regression among similar questions; treat each question as a separate test; use cronbach’s alpha
What is cronbach’s alpha?
how well each item correlates with each other
What is validity?
truth in measurement; accurately measuring what you intend to measure
construct validity
how measuring device accurately measures target construct:
- concurrent validity
- discriminant validity
concurrent validity
test correlation between two similar constructs; high correlation = good
discriminant validity
run correlation between two opposite tests of different constructs; low correlation is good
criterion validity
how well the test measures outcome external of measuring device:
1. predictive validity
- postdictive validity
- concurrent validity
predictive validity
how well the test predicts future actions
concurrent validity
two tests taken at same time to predict behavior
postdictive validity
how well the test measures behavior after occurence: known groups paradigm