Probit margins "Factor not found in list of covariates"

[Anne Klein]

Dear Stata users,

I am running a probit model with three binary indep. vars. and some other continuous dep. vars. The probit model runs perfectly, however, whenever I want to take the margings (which I only want for the three binary Independent variables), Stata tells me that the "Factor was not found in the list of covariates". I posted the log below. The variables I want the margins of (

drought_025 MMdrought_025 MM_use) are three binary indicators. What can I do to fix this?
Best, Anne

. probit hh_q14 drought_025 MMdrought_025 MM_use hhsize phoneowner wealthscore bankaccount part_assoc, vce(cluster hhid)

Iteration 0: log pseudolikelihood = -912.3496
Iteration 1: log pseudolikelihood = -904.10648
Iteration 2: log pseudolikelihood = -904.04121
Iteration 3: log pseudolikelihood = -904.04102
Iteration 4: log pseudolikelihood = -904.04102

Probit regression Number of obs = 1,585
Wald chi2(8) = 13.38
Prob > chi2 = 0.0993
Log pseudolikelihood = -904.04102 Pseudo R2 = 0.0091

(Std. Err. adjusted for 1,585 clusters in hhid)
------------------------------------------------------------------------------
| Robust
hh_q14 | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
drought_025 | -.0134295 .2133811 -0.06 0.950 -.4316488 .4047898
MMdrough~025 | -.6939054 .5643085 -1.23 0.219 -1.79993 .4121191
MM_use | -.2033974 .0956013 -2.13 0.033 -.3907726 -.0160223
hhsize | -.0233613 .0143946 -1.62 0.105 -.0515741 .0048516
phoneowner | .2052696 .0794024 2.59 0.010 .0496437 .3608956
wealthscore | .0028154 .0276162 0.10 0.919 -.0513113 .0569422
bankaccount | -.1351624 .1538211 -0.88 0.380 -.4366463 .1663215
part_assoc | .1581335 .1736723 0.91 0.363 -.182258 .4985251
_cons | -.5774606 .103155 -5.60 0.000 -.7796407 -.3752805


. margins drought_025 MMdrought_025 MM_use, post
factor drought_025 not found in list of covariates
r(322);

[Joro Kolev]

Put a toy sample of your data with -dataex- so that we can check what is going on, if what I say below is not it.

But I think Stata wants you to tell her that these are binary variables like this

. sysuse auto, clear
(1978 Automobile Data)

. summ mpg, meanonly

. gen mpgdummy = mpg>r(mean)

. probit foreign mpgdummy headroom, nolog

Probit regression                               Number of obs     =         74
                                                LR chi2(2)        =      11.28
                                                Prob > chi2       =     0.0036
Log likelihood = -39.393062                     Pseudo R2         =     0.1252

------------------------------------------------------------------------------
     foreign |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
    mpgdummy |    .712355   .3542323     2.01   0.044     .0180725    1.406638
    headroom |  -.3535888   .2334736    -1.51   0.130    -.8111886     .104011
       _cons |   .1344065   .7590577     0.18   0.859    -1.353319    1.622132
------------------------------------------------------------------------------

. margins mpgdummy
factor 'mpgdummy' not found in list of covariates
r(322);

. probit foreign i.mpgdummy headroom, nolog

Probit regression                               Number of obs     =         74
                                                LR chi2(2)        =      11.28
                                                Prob > chi2       =     0.0036
Log likelihood = -39.393062                     Pseudo R2         =     0.1252

------------------------------------------------------------------------------
     foreign |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  1.mpgdummy |    .712355   .3542323     2.01   0.044     .0180725    1.406638
    headroom |  -.3535888   .2334736    -1.51   0.130    -.8111886     .104011
       _cons |   .1344065   .7590577     0.18   0.859    -1.353319    1.622132
------------------------------------------------------------------------------

. margins i.mpgdummy

Predictive margins                              Number of obs     =         74
Model VCE    : OIM

Expression   : Pr(foreign), predict()

------------------------------------------------------------------------------
             |            Delta-method
             |     Margin   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
    mpgdummy |
          0  |    .188038   .0642029     2.93   0.003     .0622025    .3138735
          1  |   .4198893   .0923938     4.54   0.000     .2388007    .6009778
------------------------------------------------------------------------------

.

[Anne Klein]

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte hh_q14 float(MM_use drought_025 MMdrought_025)
1 0 0 0
1 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
1 0 0 0
0 0 0 0
0 0 0 0
0 1 0 0
0 0 0 0
0 0 0 0
0 1 0 0
0 0 0 0
0 1 0 0
1 0 0 0
0 0 0 0
1 0 0 0
0 0 0 0
1 0 0 0
0 0 0 0
1 0 0 0
1 0 0 0
0 0 0 0
1 1 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 1 0 0
1 0 0 0
0 0 0 0
0 0 0 0
1 0 0 0
1 0 0 0
1 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 1 0 0
0 0 0 0
0 0 0 0
1 0 0 0
1 0 0 0
0 0 0 0
0 0 0 0
1 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
1 0 0 0
0 1 0 0
0 0 0 0
0 0 0 0
0 0 0 0
1 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 1 0 0
0 0 0 0
0 1 0 0
0 1 0 0
0 0 0 0
0 1 0 0
0 1 1 1
0 0 1 0
0 0 0 0
0 1 1 1
0 0 1 0
0 0 0 0
0 0 1 0
0 1 0 0
0 1 0 0
0 1 0 0
end
label values hh_q14 hh_q14
label def hh_q14 1 "YES", modify

This is the dataex - shall I really put i. also before only binary, not categorical variables?

[Anne Klein]

Indeed, it worked, thank you!

[Joro Kolev]

You are welcome !

These are called factor variables, a bit misleadingly because they include also continuous variables and interactions. Check the help

help fvvarlist

to see how exactly they work. Basically factor variables are i.var, continuous variables are c.var, interactions are #, and all levels interactions are ##.

You use them as you need, in this case as you do not need to apply -margins- on the continuous controls, I think it is not necessary to tell Stata that the controls are continuous.

Indeed, it worked, thank you!

[Richard Williams]

As a general rule, anything to the left of the comma has to be a categorical variable, designated by something like i.var in the preceding estimation command. i.var is often a dichotomy but it can have more categories, e.g. i.relig where relig is coded 1 = Catholic, 2 = Protestant, 3 = Jewish, 4 = Muslim, etc.

As an aside, your output would be far easier to read if you used code tags. See section 12 of the Statalist FAQ. https://www.statalist.org/forums/help#stata

[Anne Klein]

Thank you, Richard and Joro. Joro, how do you interpret the marginal effect for mpgdummy = 0?