margins: factor variables may not contain noninteger values

[Lucas Kori Leonhard]

I am trying to estimate the marginal effects of my xtlogit model, which looks like this:
xtlogit onset c.l.log_welfarespending##c.l.ethnic_groups l.gdplog, re vce(robust)
The model itself runs smoothly and there are no complications at all. When I try to estimate the marginal effects:
margins log_welfarespending, at(l.ethnic_groups(= (0, 6))
I receive log_welfarespending: factor variables may not contain noninteger values. From what I have read so far sometimes Stata needs a specification on whether the variables are categorical or continuous. When I change the code to:
c.margins log_welfarespending, at(l.ethnic_groups(= (0, 6))
The error changes to only factor variables and their interactions are allowed. Any suggestions how to solve the problem?

[Clyde Schechter]

In you regression you have clearly specified log_welfarespending as a continuous variable, not a discrete one. So it simply cannot be used in -margins- the way you are trying. To get margins for specific values of that variable, you have to use the -at()- option instead.

[Lucas Kori Leonhard]

Clyde Schechter thanks a lot for the answer. Would it be possible to use the at() option twice? To display the margins
from log_welfarespending at the values 3, 4 and 5 when l.ethnic_groups are 0 and 6 ?

[Clyde Schechter]

Yes, you could do that. But there is no need to use the -at()- option for the ethnic groups: that is a discrete variable with non-negative integer values, right? You only need the -at()- option for the continuous variables. So I would do this as:

margins ethnic_groups, at(log_welfarespending = (3 4 5))

Now, that will give you the margins for all combinations of log_welfarespending = (3 4 5) and each ethnic group 0 through 6. If you only care about ethnic groups 0 and 6, the simplest thing is to just disregard the results you get for ethnic groups 1 through 5. But if that is not convenient, or if the -margins- command takes a long time to run, then you can pare it back using the double-at approach:

margins ethnic_groups, at(log_welfarespending = (3 4 5) ethnic_groups = (0 6))

[Lucas Kori Leonhard]

Clyde Schechter thank you again! My problem is the grapgh is kind of messy with all 6 levels of ethnic groups. I can edit the grapgh in a way that these levels are hidden, but due to the high standard errors of one level of ethnic groups (which I terefore want exclude from the graph) it only pictures the graph from "far away", so the 95% Cls can be shown. When tryin

margins ethnic_groups, at(log_welfarespending = (3 4 5) ethnic_groups = (0 6))

I get the error variable 'ethnic_groups' not found in list of covariates

[Clyde Schechter]

Sorry, for a couple of errors. I see that your variable is l.ethnic_groups, not ethnic_groups. Also, it shouldn't be mentioned in both places. So the command should be:

margins, at(log_welfarespending = (3 4 5) l.ethnic_groups = (0 6))

[Lucas Kori Leonhard]

The problem remains: error variable 'ethnic_groups' not found in list of covariates

[Clyde Schechter]

Post back with the exact output of your regression and the exact regression and margins commands you are using so I can see what is going on.

N.B. Important: do that by copying and pasting from your Results window or log file. Do not type it in manually, and do not edit the output in any way.