Error while using Keep option with esttab

[Akanksha Aggarwal]

Hi,

I am checking if there is heterogeneity in my results based on multiple variables.

I run the following regressions :

reg var1 var2 var3 var3#var4 var7 var8

reg var1 var2 var3 var3#var5 var9 var10

reg var1 var2 var3 var3#var6 var10 var11

I only want to keep coefficients of var2 var3 and the interaction terms
When I use keep ( var2 var3 var3#var4 var3#var5 var3#var6 ) with esttab , the error thrown is coefficient var3#var4 not found.

Could someone please help in correcting this error. Could it be because var3#var4 doesn't appear in all three regressions.

Thanks.

[Andrew Musau]

Run the regress command as

reg var1 var2 var3 var3#var4 var7 var8, coeflegend

so to see how Stata labels the coefficient. I suspect you need something like

*var3#*var4

estout is from the Stata Journal/ SSC, as you are asked to explain (FAQ Advice #12).

[Akanksha Aggarwal]

Thanks a lot, Andrew.

[Akanksha Aggarwal]

I am able to access the variables in keep now. In the regression output though using esttab, I want a single row instead ( to make the regression table compact) with coefficients of all interaction terms ( var3#var4 , var3#var5 , var3#var6 ) instead of three different rows for each of the three interactions.


How can that be done?

Thanks.

[Andrew Musau]

These are 3 separate coefficients, so you cannot include them in one row. You can run

margins, dydx(*) post

after the regression and report the average marginal effects which will not include the interactions.

[Akanksha Aggarwal]

Hi Andrew,

Thanks for the reply. The following is how I want the final regression output to look like.

	Variable 4	Variable 5	Variable 6
var 2	Coefficient	Coefficient	Coefficient
var 3	Coefficient	Coefficient	Coefficient
var 3*Variable	Coefficient	Coefficient	Coefficient

The column headings will represent the variable I am interacting variable 3 with. This will allow me to put all the three interaction terms in one row.

Thanks.

[Andrew Musau]

I cannot follow this as you are using pseudo-code to describe the model and then presenting some table based on this. Usually, the columns represent results from one model and the rows are the coefficients. 3 columns therefore imply results from 3 estimated models. Take a dataset such as Stata's auto dataset and illustrate what you want after estimating a model.