How to Interpret Coeficient as a Result of Structural Coeficient Model

[Ainun Nisa]

Hi,
Previously, I asked about my model using SEM in Stata. I already run the data and i got this result:

What I want to ask is how to interpret the coeficient in equation 3 from log_income to health condition? log_income is my independent variable and health condition is my dependent. I feel confuse because health condition variable is in chategorical form (1-4 point self assesment of health condition). Here is my data:

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(log_inc health)
17.989899 3
16.993565 3
 16.38046 2
17.175886 3
16.339638 3
 14.69098 3
 15.85413 2
16.298416 3
16.482738 3
 13.81551 2
16.993565 3
16.333206 3
16.562782 2
15.789592 4
16.300417 3
15.789592 4
 15.76142 3
16.206106 2
17.661394 2
17.843716 2
17.843716 4
18.379858 3
18.092176 3
17.399029 2
 15.60727 3
16.993565 3
16.300417 3
16.482738 3
16.906553 2
17.622173 2
16.157316 2
12.611538 3
12.611538 2
16.705883 2
16.300417 3
15.943742 3
16.993565 3
16.300417 3
17.686712 3
17.766754 4
15.894952 3
 13.81551 3
 15.76142 3
17.353567 3
 13.81551 3
 16.38046 4
 16.38046 3
 16.38046 2
 16.38046 3
16.705883 2
16.705883 4
15.943742 3
16.300417 4
16.195057 4
16.240313 3
16.357113 4
14.808763 3
17.281246 4
16.785925 3
16.157316 3
 15.60727 3
16.300417 3
16.705883 3
16.300417 4
16.300417 3
15.096444 3
16.195057 4
  16.4182 4
16.482738 2
16.077274 3
  16.4182 4
14.771022 4
 16.57552 3
16.716932 3
 15.85673 3
16.780783 3
 16.57552 3
16.993565 3
 16.58183 2
16.586473 2
 16.57514 3
14.220976 2
 15.64006 3
16.300417 3
17.111347 2
16.275099 3
16.785925 3
18.092176 3
15.687313 2
16.523561 3
17.302885 3
  16.4182 2
16.300417 2
 15.60727 2
 15.60727 2
15.201805 3
15.201805 3
16.077274 2
16.929026 3
16.993565 2
end

[Phil Bromiley]

You will increase your chances of useful answer by following the FAQ on asking questions-while you provide data, you don't provide Stata output or the model you estimated.

If health condition is categorical, then you should model it that way in GSEM - an ordinal logit probably See page 23 in the SEM documentation and elsewhere in the examples. Then you can use margins to look at changes in predicted value.

The other alternative is to treat the one to four values as continuous. In that case, a one unit increase in income results in a .0158 increase in predicted health condition. You should use margins after an estimation to help you interpret the results.