Stata Collect System: Combining two models into one model prior to making table

[Nathan Yu]

Hello,

I would like to have a table that includes only one column, but is reporting results from two models. I would think that the following command should work, however, the predictors that are shared between the two regressions do not show in the table:

webuse nhanes2l, clear
collect clear
                
collect _r_b _r_ci, tag(model[combined_results]): regress bpsystol hgb i.race i.rural
collect _r_b _r_ci, tag(model[combined_results): regress bpsystol hct i.race i.rural
collect layout (colname[hgb hct 1.race 2.race 3.race 0.rural 1.rural]#result) (model)

This results in:

. collect layout (colname[hgb hct 1.race 2.race 3.race 0.rural 1.rural]#result) (model)

Collection: default
      Rows: colname[hgb hct 1.race 2.race 3.race 0.rural 1.rural]#result
   Columns: model
   Table 1: 6 x 1

--------------------------------------
                  |   combined_results
------------------+-------------------
Hemoglobin (g/dL) |                   
  Coefficient     |           1.714797
  95% CI          | 1.384949  2.044645
Hematocrit (%)    |                   
  Coefficient     |           .6554286
  95% CI          | .5325924  .7782647
--------------------------------------

Although I understand that I can use the following code with two separate models, one for each regression, however, this would result in a table with two columns:

webuse nhanes2l, clear
collect clear
                
collect _r_b _r_ci, tag(model[model_1]): regress bpsystol hgb i.race i.rural
collect _r_b _r_ci, tag(model[model_2]): regress bpsystol hct i.race i.rural
collect layout (colname[hgb hct 1.race 2.race 3.race 0.rural 1.rural]#result) (model)

Which results in:

. collect layout (colname[hgb hct 1.race 2.race 3.race 0.rural 1.rural]#result) (model)

Collection: default
      Rows: colname[hgb hct 1.race 2.race 3.race 0.rural 1.rural]#result
   Columns: model
   Table 1: 19 x 2

-----------------------------------------------------------
                  |             model_1             model_2
------------------+----------------------------------------
Hemoglobin (g/dL) |                                        
  Coefficient     |            1.714797                    
  95% CI          |  1.384949  2.044645                    
Hematocrit (%)    |                                        
  Coefficient     |                                .6554286
  95% CI          |                      .5325924  .7782647
White             |                                        
  Coefficient     |                   0                   0
Black             |                                        
  Coefficient     |            5.155836            4.643728
  95% CI          |  3.642232  6.669441  3.146876   6.14058
Other             |                                        
  Coefficient     |            -1.63656           -1.674051
  95% CI          | -4.891694  1.618573 -4.928377  1.580275
Urban             |                                        
  Coefficient     |                   0                   0
Rural             |                                        
  Coefficient     |            1.331342            1.325153
  95% CI          |  .3859748  2.276709  .3800328  2.270272
-----------------------------------------------------------

I understand that the covariates result in different coefficients and confidence intervals pending on the models, however, with the actual data I am using (I am not using nhanes2l data), the coefficients/CI's are exactly the same in the two models that I want to include in one column.

How can I make a table that includes the coefficient/CI's for hematocrit in the first table, that contains the results for model 1?

Thank you in advance!

[Simon Pfaff]

Hello Nathan,
maybe you are looking for something like this?

I am sure you could further format the table.
I wasn't able to get rid of the labels "model_1" & "model_2"...

webuse nhanes2l, clear
collect clear

collect get, tag(model[model_1]): regress bpsystol hgb i.race i.rural
collect get, tag(model[model_2]): regress bpsystol hct i.race i.rural

collect layout (colname[hgb 1.race 2.race 3.race 0.rural 1.rural]#model[model_1]#result[_r_b  _r_ci] colname[hct]#model[model_2]#result[_r_b  _r_ci])

collect style showbase off
collect style cell, nformat(%5.2f)
collect style column, extraspace(1)
collect preview

--------------------------------
Hemoglobin (g/dL) |             
  model_1         |             
    Coefficient   |         1.71
    95% CI        |   1.38  2.04
Black             |             
  model_1         |             
    Coefficient   |         5.16
    95% CI        |   3.64  6.67
Other             |             
  model_1         |             
    Coefficient   |        -1.64
    95% CI        |  -4.89  1.62
Rural             |             
  model_1         |             
    Coefficient   |         1.33
    95% CI        |   0.39  2.28
Hematocrit (%)    |             
  model_2         |             
    Coefficient   |         0.66
    95% CI        |   0.53  0.78
--------------------------------

[Nathan Yu]

Thank you very much Simon, this helps me understand how to frame tables using the Collect system! However, I would need models to be present on the columns still. I plan to eventually combine 6 models into three columns, where only one coefficient/confidence interval from one model is appended to a different model. Is there is a way to add/append just one set of coefficient/confidence interval from one model to the tags associated with a separate model, so that ultimately the resultant table will include estimates from two separate model in the same column?

Best regards,

[Simon Pfaff]

Hello Nathan,

I hope I don't get you wrong, "collect remap" might help.

webuse nhanes2l, clear
collect clear

collect get, tag(model[model_1]): regress bpsystol hgb i.race i.rural
collect get, tag(model[model_2]): regress bpsystol hct i.race i.rural

collect remap model[model_2]=model[model_1], fortags(colname[hct])

collect layout (colname[hgb hct 1.race 2.race 3.race 0.rural 1.rural]#result[_r_b  _r_ci]) (model[model_1])

collect style showbase off
collect style cell, nformat(%5.2f)
collect style column, extraspace(1)
collect preview

--------------------------------
                  |      model_1
------------------+-------------
Hemoglobin (g/dL) |             
  Coefficient     |         1.71
  95% CI          |   1.38  2.04
Hematocrit (%)    |             
  Coefficient     |         0.66
  95% CI          |   0.53  0.78
Black             |             
  Coefficient     |         5.16
  95% CI          |   3.64  6.67
Other             |             
  Coefficient     |        -1.64
  95% CI          |  -4.89  1.62
Rural             |             
  Coefficient     |         1.33
  95% CI          |   0.39  2.28
--------------------------------

[Nathan Yu]

Hi Simon,

Thank you so much -- this is exactly what I needed!!

Best regards,