Hello, I assume a nonlinear relationship between TobinsQ and sumholdingpct.


Is it then appropriate to apply variable transformation on both sides?

I thought about logarithm transformation for the dependent variable and also including squared variables for the main independent variable of interest at the same time?
Or should i just use a transformation on the independent variable of interest?

Something like this:

reg ln(TobinsQ) SumHoldingPct SumHoldingPct^2 DebtTotalAssets Dividend FirmSize QuickRatio SalesGrowth i.year i.industry i.country, robust cluster(industry)