I want to asses impact of fertilizer, seeds and irrigation on crop productivity. Total observation number is 100 and its a panel model. The issue is that while dependent variable requires log transformation, independent require either normal or square root or square transformation to attain normality. Does this kind of a model with different transformations make sense? And how can I integrate bootstrapping with panel model so as to avert the issue of small sample size?
You don't need to transform your independent variables. I even don't transform the dependent variable unless I would have to. Make sure the residuals are normal.
do not transform
Article TO DETERMINE SKEWNESS, MEAN AND DEVIATION WITH A NEW APPROAC...
First of all residual analysis is necessary before any transformation is applied in regression modeling. In the situation of detection of nonlinearity, transformation in the dependent variable can be done, if it results in desired correction in residual analysis. Transformations for independent variables is unnecessary unless such functional forms is found very suitable in explaining association between variables.
- Badges
- Science topic