Wednesday, November 5, 2008

write up

For question 1A, we are asked to report whether adding more predictors will improve the predictability of all of the X variables or not. Do we have to report the R square for each model or do we have to report the R square for the overall model only? When the change of R square between two models is 0.00, does it mean adding extra predictor variable(s) does not improve the prediction? Thanks!

2 comments:

Mari said...

Unless you report the R square change for a block, you have no justification for deciding whether or not the prediction significantly improved.

Unless you report the final model statistics, your reader has no way of knowing whether or not the final model was significant.

Unless you report the initial step statistics, your reader has no way of knowing if your initial model were significant or not.

Mari said...

An R2 change of .00 means that the addition of that predictor did not change R2 more than .004. This value must be interpreted along with its F & p. If not significant, then yes, you would conclude that the addition of the predictor(s) on that step did not significantly add to prediction of the outcome variable.