xtqreg is not estimating coefficients of category variable

[ajay pasi]

My data is assessment data of a college. Where each student was assessed in 16 courses. So, each student ID repeats 16 times in the dataframe. I want to study category of the student has effect on performance.\

Steps I followed:

1. xtset student_unique_ID
2. Ran xtqreg. But I encountered the coefficients of category variable were omitted.

. xtqreg nblind75_blind_diff   i.course1 ib4.caste_minority_mix_new_NUM_new i.T_nature i.T_gender i.semester attendence_percent founda
> tion_common i.T_jati_new



                              MM-QR regression results
Number of obs = 9769
.5 Quantile regression
------------------------------------------------------------------------------------------------
                               | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------------------------+----------------------------------------------------------------
                       course1 |
                          eco  |          0  (omitted)
                          eng  |          0  (omitted)
                        hindi  |          0  (omitted)
                      history  |          0  (omitted)
                        maths  |          0  (omitted)
                          pol  |          0  (omitted)
                          skt  |          0  (empty)
                               |
caste_minority_mix_new_NUM_new |
                            1  |          0  (omitted)
                            2  |          0  (omitted)
                            3  |          0  (omitted)
                            5  |          0  (omitted)
                               |
                    2.T_nature |  -1.082085   .3321962    -3.26   0.001    -1.733178   -.4309927
                    2.T_gender |  -.0273043   .4535507    -0.06   0.952    -.9162472    .8616387
                               |
                      semester |
                            0  |          0  (empty)
                            1  |  -1.389002   .0150841   -92.08   0.000    -1.418566   -1.359438
                            2  |  -.4392321   .5603047    -0.78   0.433    -1.537409    .6589449
                            3  |   .4923732   .4811123     1.02   0.306    -.4505896    1.435336
                            4  |   2.936472   .5412501     5.43   0.000     1.875641    3.997302
                            5  |          0   .5285921     0.00   1.000    -1.036021    1.036021
                            6  |          0  (empty)
                               |
            attendence_percent |   .0888965   .0113053     7.86   0.000     .0667385    .1110545
             foundation_common |          0  (omitted)
                               |
                    T_jati_new |
                            2  |  -.4875162   .4249853    -1.15   0.251    -1.320472    .3454396
                            3  |  -1.547974    .435074    -3.56   0.000    -2.400703   -.6952445
------------------------------------------------------------------------------------------------


.

caste_minority_mix_new_NUM_new is the variable for students' category.


Can anyone guide what is happening, and what am I doing wrong?

[Joao Santos Silva]

Dear ajay pasi,

If I understand correctly, the variables being dropped are constant for each student, and therefore collinear with student fixed effects. Such variables are dropped in any model/estimator that includes student fixed effects.

Best wishes,

Joao

[ajay pasi]

Dear Joao Santos Silva

You are right that category of students repeat becaue it naturally can't change.

The coefficients of categorical variable at differnt quantiles of depVar, are primary for my research. What appraoch should I go for?

bsqreg ?

or xi: qreg2 ...., cluster (student ID)?

If it is possible to do estimation of coefficients (of categorical variable)with precision in my case?

[Joao Santos Silva]

You can use qreg2 with clustered standard errors, but you cannot use any model/estimator with fixed effects.

[ajay pasi]

Thank you for your insights, Joao Santos Silva

I noticed that when I use `xtreg`, it does report the coefficient for the category variable, even when category variable is constant for shudent IDs. However, when I switch to `xtqreg`, it doesn't estimate the coefficient for this category.


Additionally, I found that when I transformed the dependent variable into a dummy variable, `xtlogit` was able to report the coefficients for this category variable.