I would like to do regression using residuals (as an adjustment for total energy (as suggested by Willett et al, 1997) for one variable and I would like to do multiple imputation also for some other variables with missing values. Which one I should do first ? The final model will include the residual variable and the variable from result of multiple imputation as explanatory variables.
What type of regression are u doing? I have done linear multiple regression analysis . I can help u with that if u want. There are basically three types of regression Bivariate regression, Stepwise Regression and Linear Multiple Regression Analysis. And for that R software is very helpful. So if u tell be the type of equation may be i can help u with that. I will also give u the contact of renowned statistician in my university. if u want.
Thanks
Ghazala Perveen Khan
Freckleton (2002) cautions against regression using residuals. I have actually never heard of this technique--not used in my discipline. I have orthogonalized covariates that have some collinearity before. And I have done various forms of multiple imputation. R has a mice() package for multiple imputation using chained equations. There are certain kinds of regression models (e.g., HLM random effects) that take missing values--provided that you have the right data structure.
http://onlinelibrary.wiley.com/doi/10.1046/j.1365-2656.2002.00618.x/full
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression