Is OLS regression using a bootstrapping approach adequate for low base-rate variables? Is there any practical alternative, I mean, using the SPSS?
Could you provide more information? Low base rate variables suggests that the data might be discrete in which case some kind of generalised linear model might be more appropriate, however modelling sparse data can be very tricky.
Thank you! No, they are continuous variables, five predictors and one continuous dependent variable, but one of the predictors and the dependent variable are very positively skewed.
OK. Positive skew isn't necessarily too difficult to deal with. Generally I would look to transform the variables for a regression model. Often the first approach is to consider a log transform of the skewed variables but there are other approaches.
It is a form of normalizing variables. Just one more question, it would probably better to transform all variables, in this case, all precitors and the criterion variable, even if some are not skewed?
No - you would only transform the skewed variables. Also to check the model look at the distribution of residuals rather than the raw data.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS