Stata exclude values in a glm regression

[Soren Frambo]

Hello. I have a problem where stata excludes values.

I am running the following code:

glm positiv_pleural_sample ib0.d_d_inden_for_90_dage i.COPD i.CKD i.IHD_CHF i.DM i.age_group i.respirator_support i.circulatory_support i.renal_support i.gender_num i.ab_spektrum i.bloddyrkning i.luftvejssekret, link(log) family(poisson) vce(robust) eform

The result for age_group is missing age_group 1. The result i post is only showing age_group.

----------------------------------------------------------------------------------------
| Robust
positiv_pleural_sample | IRR std. err. z P>|z| [95% conf. interval]
-----------------------+----------------------------------------------------------------

|
age_group |

2 | .5104993 .3450695 -0.99 0.320 .1357156 1.920262
3 | .2002841 .1243218 -2.59 0.010 .0593311 .676099
4 | .1381519 .0931424 -2.94 0.003 .0368531 .5178923

----------------------------------------------------------------------------------------

age_group is coded as such:
replace age_group = 1 if alder_ved_dræn >= 18 & alder_ved_dræn <= 25
replace age_group = 2 if alder_ved_dræn >= 26 & alder_ved_dræn <= 50
replace age_group = 3 if alder_ved_dræn >= 51 & alder_ved_dræn <= 75
replace age_group = 4 if alder_ved_dræn >= 76 & alder_ved_dræn <= 100

I dont have any missing values in age_group. I got a total of 13 sampels in the age_group 1.

What can i do so age_group 1 is not excluded.

Thanks in advance.

[Joseph Coveney]

What can i do so age_group 1 is not excluded.

If there weren't any warning messages at the beginning of iterations that the first category of the age group variable is omitted for some reason, then something like the following should work.

glm positiv_pleural_sample . . . ibn.age_group . . . , noconstant link(log) family(poisson) vce(robust) eform

[Andrew Musau]

You will notice that is true for all categorical variables (indicated as "i.catvar" using factor variable notation). You can make one exception, as Joseph shows in #2, but that is at the expense of excluding the intercept. See the discussion pertaining to the dummy variable trap in https://en.wikipedia.org/wiki/Dummy_...e_(statistics).

[Maarten Buis]

That is as it should be. 1 is the reference category for age group. So the coefficients are a comparison of its age-group and the reference age group. So 2.age_group compares the 26 to 50 year olds with the 18 to 25 year olds, and 3.age compares the 51 till 75 year olds to the 18 to 25 year olds. Now it does not make sense to include 1.age: that would compare the 18 till 25 year olds with the 18 till 25 year olds. We don't need to estimate that difference: it is 0 by definition.

[Soren Frambo]

Thank you all.

I will analyse the above recommendations.