note: continuous option implied because a factor with only one level was specified in the eydx() option

[Osiris Parcero]

I ran a ppmlhdfe regression on an unbalanced panel data looking at the citations received by individual sole-authored papers over time.

ppmlhdfe exclud_self   i.paper_age   i.dead_dumyA1    i.dead_dumyA1#c._2Year_citeAtDeath  i.dead_dumyA1#c.AuthorActiveAtDeathRel 1.dead_dumyA1#c._2Year_citeAtDeath#c._2Year_citeAtDeath, absorb(i.paper)  vce(cluster paper author_id) d

Then, to obtain the semielasticity, I run the following command, where dead_dumyA1 is a factor variable having two levels: equal to 1 if the author of a paper is dead in the particular year and 0 otherwise.

margins, eydx(i.dead_dumyA1)

As shown by the output above, the command runs and provides me with the semielasticity, but I am worried about the following warning message:
“(note: continuous option implied because a factor with only one level was specified in the eydx() option)”
I cannot understand this message. Why is Stata saying that this factor variable has only one level?
I carefully checked this variable for the sample used in this regression, and there is no individual for which dead_dumyA1 is equal to 0 in all the years or all equal to 1 in all the years.
I am worried because a “continuous option implied” means that Stata is calculating a derivative, but I need the calculation of the discrete difference because " dead_dumyA1” is a factor variable with two levels.
I am using Stata MP 14.2.
Any help is greatly appreciated.

[Osiris Parcero]

Later today I replicated the same regression followed by the same margins’ command but with only 9 papers (individuals) and 204 observations. In this new case, I am even surer that the dead_dumyA1 has both 0s and 1s for each single individual. However, the margins command continues giving the message:
“(note: continuous option implied because a factor with only one level was specified in the eydx() option)”

[Andrew Musau]

ppmlhdfe is from SSC (FAQ Advice #12). Look at your estimation command:

ppmlhdfe exclud_self i.paper_age i.dead_dumyA1 i.dead_dumyA1#c._2Year_citeAtDeath i.dead_dumyA1#c.AuthorActiveAtDeathRel 1.dead_dumyA1#c._2Year_citeAtDeath#c._2Year_citeAtDeath, absorb(i.paper) vce(cluster paper author_id) d

You are using "dead_dummy" both as a factor variable and a continuous variable. How do you expect margins to calculate the semi-elasticity? Change "1.dead_dumyA1" to "i.dead_dumyA1".

[Osiris Parcero]

Thank you so much Andrew for your precise and useful answer, as usual. This solves the problem.