What is the acceptable values for multiple linear regression assumptions including
R square
Adjusted R square
Standard error of estimate
Linear correlation between dependent and independent variables
Linear correlation between independent variables
VIF
Cooks distance
This question can not be answered. R^2 and adj. R^2 and correlations between the variables are NOT assumptions, but the result of your analysis. There can be no agreement about their meaningful magnitude across research areas. Whereas in social sciences (and only subfields of it) an R^2 of .30 is considered high, in physics this would clearly indicate that something is wrong. Therefore, you have to look into publications in your field, which magnitude you can expect and should be reached to be meaningful.
For Cooks i and and VIF, there are only rules of tumb. VIF should be smaller than 10 and for Cooks i there is a rough rule to be < 1 or a bit more strict and conservative it should be < 4/N (Gordon, 2010), but which leads to very small boundaries in large samples.
VIF and Tolerance (not Cook's distance) are opposite measures of multicollinearity. The first must be more than 1.00 and the second less than 1.00
Your regression model depends on if it is significant or not (ANOVA) not on how many % of variance R2 explains.
Your regression model linear model or logistic model is does not depend on R^2 value rather on the significance of the variables. But if the R^2 value may improve by Imputation of NA values or by detecting and taking remedial of multicollinearity etc.
dear: There are no specific values for all the estimates and tests that you mentioned ... and what determines the values are several factors, including sample size, homogeneity of data, data distribution, the number of independent variables that affect the phenomenon under study and there are several other influences.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size