Yes, of course.

https://stats.stackexchange.com/questions/431769/can-i-complete-a-moderation-analysis-with-transformed-data

Anouk:

Natural log transformation is very commonly used to handle heteroskedasticity issues. The most common is the log transformation, e.g.,

lnx =a*+b*lny+c*lnz +e*

Occasionally, a semi-log transformation may be used, this is the case when there are negative values in the variable, e.g., let's say variable z contains negative values, so you can try:

lnx =a*+b*lny+cz +e*.

Alternatively, you can also try to transform variable z into all positive values such as z* = (1+z), when the negative values are not big, such as percentages, e.g., -10% = -0.10.

The problem here is that it's difficult to interpret the estimated coefficient on z or lnz.

Finally, please note very carefully that after log transformation, the interpretation of the coefficients are very different. The estimated coefficients in the log transformed equation are now elasticities.

I hope this helps.