predictor variables should not have a variance of zero
10 of 23
independence
all outcome variables should be independent
11 of 23
linearity
relationship between predictors and outcomes is linear
12 of 23
No perfect multicollinearity
correlation between predictor variables cannot be too strong
13 of 23
homoscediasticity
at each level of predictor variable, the residual terms should be constant
14 of 23
independent errors
for any two data points the residual points should not correlate
15 of 23
Normally distributed errors
residual values are random and normally distributed.
16 of 23
How to check assumptions
**- calculate in advance
VT- check measurements
NZV- check by calcluating SD of variables
I- outcomes scores all from different people
L- SP**
NPM- check VIF
H-SP**
IE- durbin watsin
NDE- SP**
17 of 23
VIF stands for...
Variance Inflation Factor
18 of 23
Results of VIF mean...
VIF> 1 = biased regression
VIF> 10 = definately a problem
19 of 23
results of tolerance ...
Below 0.1 = serious problem
below 0.2 = potential problem
Comments
No comments have yet been made