Problem with using -margins- command without using -fvvarlist-

[Mihir Sharma]

Hello,

I am working with a smal panel dataset (N=20, T=20) on Stata 15. I first estimate an equation like this:
y = a + b₁x₁ +b₂x₂ + b₃x₁*x₂ (Eq. 1) where x₁ and x₂ are continuous variables and b₁, b₂, b₃ are estimated coefficients.

I then want to use the margins command to compute the marginal effect of x₁ on y as x₂ varies. If I use, say xtreg to estimate equation 1, I can easily do this using the following code. Since I use fvvarlist to write the interaction terms, margins know that two continuous variables are being interacted. When it computes dydx for x₁, as x₂ varies, it takes this into account and I get different values of the marginal effect for each value of x₂ between 0 and 1 (see code below).

xtreg y c.x1##c.x2, fe r
margins, dydx(x1) at(x2=(0(.01)1))

However, I am using a user-written command (xtivreg2) to estimate equation 1. xtivreg2 does not allow using fvvarlist (I get the following error if I try "cx1#c: operator invalid".)

Is it possible to (manually?) compute the marginal effect of the variable without depending on fvvarlist? E.g., could I replicate the code above without defining interaction terms using fvvarlist?

Many thanks,
Mihir

[Phil Bromiley]

With nice linear models, you can easily calculate the derivatives from the original parameters. If
y = a x1 + b x2 + c x1*x2
dy/dx1= a + c x2
If you want statistics some of the test routines will do them for you.

[Mihir Sharma]

Thanks Phil. To generate confidence intervals, could you please suggest some test routines that I could use?