Lec 8 Flashcards
what is a linear regression component
one predictor (any level of data)
one I/R outcome
multilinear regression components
multiple predictors (any level of data)
ONE I/R outcome
logistic regression components
1 or more predictors (any level)
one categorical outcome (2 levels only!)
multinominal logistic regression components
one more more predictors (any level)
one categorical outcome (multi levels)
ordinal logistic outcome components
one more more predictors (any level)
one ordinal outcome (multi levels)
T/F: the unstandardized constant should always be reported for linear regressions
T
assumptions for linear regression
data must be linear
normality
homoscedasticity
free of influential outliers
independent data
what stat test is for normality?
Shapiro-Wilk (want to be not significant)
what stat test is for outliers?
Cook’s distance
issue with Shapiro-Wilk test
designed for very large populations
what value do you want the skewness/kurtosis to be?
<+2 or >-2
stat test for HOV
Levene’s test (want to be not significant)
how to test for homoscedasticity
scatterplot
what is homoscedasticity?
variance of outcome is same at all levels of predictor
what value for Cook’s distance is a problem?
> +1
what is the Durbin-Watson test for?
independence of observation
Durbin-Watson test values range from ___ to ____ and ___ is perfect
0-4
2 is perfect
what is bad for the Durbin-Watson test?
outside of 0-4
what new assumption must be met for multiple regression?
multicollinearity (do not want it)
what is considered a high correlation that violates multicollinearity?
> 0.9
how to test for multicollinearity?
VIF (variance inflation factor)
Tolerance (TOL)
VIF (variance inflation factor) should be <___ to meet multicollinearity assumption
<10
Tolerance (TOL) should be >___ to meet multicollinearity assumption
0.1
T/F: just because a predictor is significant in a model doesn’t mean its significant alone
T
