plotting conditional predicted mean and CI

[Simona Gamba]

Hi.

I have the following model:
xtreg ln_price ratio ratio2 ratio3 ln_gdp ln_export, fe cluster(prod)
(where fixed effects are at the product level, ratio2 is the squared value of ratio, and ratio3 is the cubic value of ratio).

I would like to plot the relationship between "ratio" and "ln_price", where the predicted value of "ln_price" is computed at the average value of covariates ("ln_gdp" and "ln_export"). I would also like to plot the confidence intervals.

Since I was not able to find a Stata command to compute the conditional predicted mean, what I am doing now is the following:

xtreg ln_price ratio ratio2 ratio3 ln_gdp ln_export, fe cluster(prod)
matrix list e(b)
scalar cons=_b[_cons]
scalar b1=_b[ratio]
scalar b2=_b[ratio2]
scalar b3=_b[ratio3]
scalar a1=_b[ln_gdp_pc_e]
scalar a2=_b[ln_export_e]
egen ln_gdp_mean= mean(ln_gdp_pc_e)
egen ln_export_mean= mean(ln_export_e)
gen yhat= cons+b1*ratio +b2*ratio2 +b3*ratio3 +a1*ln_gdp_mean +a2*ln_export_mean
twoway (line yhat ratio)

1) Is there a Stata command to compute the conditional predicted mean? I am using Stata14.

2) How could I compute the confidence intervals in this case?

[Fei Wang]

Simona, here is an example of plotting the predicted "ln_wage" against various values of "tenure", at the mean value of "hours".

webuse nlswork, clear
xtreg ln_wage tenure hours, fe vce(cl idcode)

margins, at(tenure=(0(5)25))
marginsplot