New on SSC: -aextlogit- Average elasticities for fixed effects logit

[Joao Santos Silva]

With the usual thanks to Kit Baum, aextlogit is now available on SSC.

aextlogit is a wrapper for xtlogit which estimates the fixed effects logit and reports estimates of the average (semi-) elasticities of Pr(y=1|x,u) with respect to the regressors, and the corresponding standard errors and t-statistics. The method used to compute the (semi-) elasticities was first described by Kitazawa (2012); see Kemp and Santos Silva (2016) for further details.

Please do let me know if you have problems with the files.

Best wishes,

Joao

[nokya oppong]

hello Mr. Silva, i'm new to stata and i'm trying to run an mlogit model on a dataset. i also need to save a do file. please can you help me out. thanks

[Linh Nguyen]

Joao Santos Silva: Hi Joao,

Following the sample in -help aextlogit-, I have the following results:

. webuse union, clear
(NLS Women 14-24 in 1968)

. aextlogit union age grade i.not_smsa south##c.year
note: multiple positive outcomes within groups encountered.
note: 2,744 groups (14,165 obs) dropped because of all positive or
      all negative outcomes.

Iteration 0:   log likelihood = -4516.5881  
Iteration 1:   log likelihood = -4510.8906  
Iteration 2:   log likelihood =  -4510.888  
Iteration 3:   log likelihood =  -4510.888  

Conditional fixed-effects logistic regression   Number of obs      =     12035
Group variable: idcode                          Number of groups   =      1690
                                                Obs per group: min =         2
                                                               avg = 7.1213018
Log likelihood  = -4510.888                                    max =        12

                   Average (semi) elasticities of Pr(y=1|x,u)
------------------------------------------------------------------------------
       union |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |   .0553283   .0747497     0.74   0.459    -.0911784    .2018351
       grade |   .0635102   .0326132     1.95   0.051    -.0004106     .127431
  1.not_smsa |   .0174948   .0880763     0.20   0.843    -.1551315    .1901211
     1.south |  -2.222936   .5265615    -4.22   0.000    -3.254978   -1.190895
        year |  -.0495603   .0753108    -0.66   0.510    -.1971668    .0980462
             |
south#c.year |
          1  |   .0205552   .0064763     3.17   0.002      .007862    .0332484
------------------------------------------------------------------------------
Average of union = .22179389 (Number of obs = 26200)

My explanation about average (semi) elasticities:

When age increases 1%, the probability of participating in union creases 5.5%

Do I explain average (semi) elasticities correctly?

I would really appreciate all the help I can get.

Best regards

[Joao Santos Silva]

I am afraid that is not the interpretation of a semi-elasticity. The right interpretation is that when age increases one year, on average the probability of being unionised goes up by 5.5%.

Best wishes,

Joao

[Yotam Shmargad]

Hello Joao - is there a package equivalent to aextlogit for xtpoisson?

[Joao Santos Silva]

Dear Yotam Shmargad

For xtpoisson that in not really needed because the coefficients have a natural interpretation as (semi-) elasticities.

Best wishes,

Joao

[Yotam Shmargad]

That's helpful, really appreciate your reply Joao.

[Anna Petrova]

Dear Joao,

I am trying to use the recommended command by you - aextlogit - as a replacement to the fixed effect login model to interpret the coefficients somehow.

Before I had the coding like:

xtset V2
xtlogit charity_participation i.generation sex i.income_class i.educ_level happiness char_conf marital_status i.religion, fe

where V2 is country

If I want to use aextlogit how should my command to run the fixed effect logit model look like:

xtset V2
xtlogit charity_participation i.generation sex i.income_class i.educ_level happiness char_conf marital_status i.religion, nolog

I run it and received the results but not sure that I have done it correctly as this command is absolutely new for me:

Thanks,
Anna

[Joao Santos Silva]

Dear Anna,

What you are doing looks OK to me.

Best wishes,

Joao

[Anna Petrova]

Dear Joao,

Thanks!

[Anna Petrova]

Dear Joao,

I have a question for you regarding the usage of command margins in order to see modeled adjusted rates of categorical variable. On the photo above you can see my model. And I want to use

margins generation

Do you know if it is allowed to use this command with aextlogit model and if yes does it have the same interpretation when I use it in ismple logit model with FE?

[Joao Santos Silva]

Dear Anna Petrova

I am afraid you cannot use the margins command after estimating the fixed effects logit; that is the point of having the aextlogit command; you also cannot use margins after this command.

Best wishes,

[Anna Petrova]

Joao Santos Silva

Thank your for your answers. The problem that I have received some input by running margins command after aextlogit model. So I was a bit confused.

[Joao Santos Silva]

Dear Anna Petrova,

You are right, Stata should not allow that; the results you get are rubbish!

Best wishes,

Joao

[Valeria Fernanda]

Dear @Joao Santos Silva

I used the command aextlogit and I get coefficients higher than 1 in absolute terms, how should I interpret this in case of a categorical variable (10 categories)?, here is what I did:



. aextlogit runhappy i.hours_work rdegree rfemale rblack rnotmar i.age_work i.slfrep_health i.tot_income i.t
note: multiple positive outcomes within groups encountered.
note: 10450 groups (26788 obs) dropped because of all positive or
all negative outcomes.
note: rdegree omitted because of no within-group variance.
note: rfemale omitted because of no within-group variance.
note: rblack omitted because of no within-group variance.

Iteration 0: log likelihood = -3038.884
Iteration 1: log likelihood = -3021.2335
Iteration 2: log likelihood = -3021.206
Iteration 3: log likelihood = -3021.206

Conditional fixed-effects logistic regression Number of obs = 8603
Group variable: id Number of groups = 2206
Obs per group: min = 2
avg = 3.8998187
Log likelihood = -3021.206 max = 11

Average (semi) elasticities of Pr(y=1|x,u)
-------------------------------------------------------------------------------
runhappy | Coef. Std. Err. z P>|z| [95% Conf. Interval]
--------------+----------------------------------------------------------------
hours_work |
2 | .1056477 .1419757 0.74 0.457 -.1726196 .3839151
3 | .1321843 .1620739 0.82 0.415 -.1854747 .4498434
4 | .1560204 .1613092 0.97 0.333 -.1601398 .4721805
5 | .3338053 .1941698 1.72 0.086 -.0467605 .7143711
6 | .0978994 .1836477 0.53 0.594 -.2620434 .4578422
7 | .3805554 .2975376 1.28 0.201 -.2026075 .9637184
8 | .2739321 .2790611 0.98 0.326 -.2730176 .8208818
9 | .9144745 .3207676 2.85 0.004 .2857816 1.543167
10 | -1.047197 .5083983 -2.06 0.039 -2.043639 -.0507548


Many thanks in advance