log transformation with spss ln() function before or after Multiple Imputation?

My data is both positively and negatively skewed. I am supposed to use the ln() log function in SPSS to cope with the skewness. Also, I have to use the "multiple imputation" function in spss to cope with missing data.

Now I my questions:
1. Should I use the log function before or after I imputed the data? Intuitively, I think it would make more sense to use the ln command before imputing the data?
2. To use the ln command on variables that also include 0- values, I plan to transform all values of the variable with X + 1. Should I do it with all variables or only with the ones that have 0-values?
3. What indicates whether I should use linear regression or predictive mean matching as model type for scale variables?

Last edited: