Hi all,

I have two questions I was hoping to get some help:

1) I have a panel(longitudinal) model of the following form: Yit= b0 + b1X1it + b2ln(X2it) + b3X1it * ln(X2it)+ b4Zit where both X1 (globalization measure) and X2 (per capital military spending) are continuous variables and ln(X2it) is the logarithmic transformation of X2. The slope of X1 would be b1+b3*ln(X2it) and ln(X2 ranges from (-3.41) to (9), b1= -0.008 and b3=0.002. Therefore, for ln(X2)=-3.41 the marginal effect of X1 would be -0.008 + 0.002*(-3.41) = - 0.0148. Would I interpret this as a globalization index increases by 1 unit Y on average decreases by - 0.0148 units (holding all else equal)? In other words does it matter that X2 is ln transformed? Additionally, if now I was to interpret the marginal effect of X2 which would be equal to b2 + b3*X1 ( for a value of X1=10 the summ would be 0.3) would I say for a 1 percent increase in X2, on average Y will increase by 0.3*0.001=0.003 units i.e., just a regular linear-log interpretation?

2) In one of my models I am using xtpcse and the regression model is the same i.e. Yit= b0 + b1X1it + b2ln(X2it) + b3X1it * ln(X2it)+ b4Zit. When I run " margins, at (ln(X2)=(-3.41(1)9)) dydx(X1) vsquish I get (not estimable) output which only happens with only one of my three dependent variables. I was wondering what might cause this kind of output.

Thank you