gologit2 output interpretation

[Vlad Alexandru]

Hello!

I am a beginner in Stata...
While trying to analyze self assessed health determinants on a relatively large sample of office workers, after using the ordered logit and Brant test, I used gologit2, since the parallel assumption was violated.
I found little information on interpreting the actual results ( most relevant here - https://www.researchgate.net/profile...-variables.pdf ).

However, I am not sure i correctly understood it or what i did wrong because my coefficients seem to have no logic.
Variables noise, envir, safe, holiday, meal, expense alll are 0,1 (0 = the state they denote does not exist, 1= it exists; ie. safe =0 if the neighbourhood is not safe and 1 if it is safe, or expense =0 if the subject cannot cover unexpected expense and expense =1 if they can cover unexpected expense).

The main thing i do not understand is how most of the variables can be negative, including the ones that should explain a positive effect if value is 1.

Below you can see the regression output.

gologit2 health age male noise envir safe holiday meal expense part full, auto(.025) lrf store(gologit2)
Generalized Ordered Logit Estimates Number of obs = 54,145
LR chi2(28) = 6402.41
Prob > chi2 = 0.0000
Log likelihood = -55497.453 Pseudo R2 = 0.0545

------------------------------------------------------------------------------
health | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
1 |
age | .0458698 .0008758 52.38 0.000 .0441533 .0475862
male | -.2364328 .0172864 -13.68 0.000 -.2703135 -.2025521
noise | -.1742681 .028502 -6.11 0.000 -.230131 -.1184052
envir | -.0344714 .0312741 -1.10 0.270 -.0957676 .0268248
safe | -.1966342 .0272342 -7.22 0.000 -.2500122 -.1432562
holiday | -.2397132 .0273712 -8.76 0.000 -.2933597 -.1860667
meal | -.1411564 .0402458 -3.51 0.000 -.2200368 -.0622761
expense | -.3773865 .0252299 -14.96 0.000 -.4268361 -.3279369
part | .0844039 .032096 2.63 0.009 .0214969 .147311
full | .2146068 .0267904 8.01 0.000 .1620985 .267115
_cons | -.2366396 .0660139 -3.58 0.000 -.3660245 -.1072548
-------------+----------------------------------------------------------------
2 |
age | .0560607 .0011065 50.67 0.000 .0538921 .0582293
male | -.2364328 .0172864 -13.68 0.000 -.2703135 -.2025521
noise | -.2684869 .0319313 -8.41 0.000 -.3310711 -.2059027
envir | -.2389496 .0344651 -6.93 0.000 -.3064999 -.1713994
safe | -.1966342 .0272342 -7.22 0.000 -.2500122 -.1432562
holiday | -.5363414 .0297612 -18.02 0.000 -.5946722 -.4780106
meal | -.1411564 .0402458 -3.51 0.000 -.2200368 -.0622761
expense | -.575941 .0284424 -20.25 0.000 -.6316872 -.5201949
part | .0844039 .032096 2.63 0.009 .0214969 .147311
full | .0136818 .0311533 0.44 0.661 -.0473775 .074741
_cons | -2.504754 .0757812 -33.05 0.000 -2.653282 -2.356225
-------------+----------------------------------------------------------------
3 |
age | .0555251 .0027324 20.32 0.000 .0501698 .0608804
male | -.2364328 .0172864 -13.68 0.000 -.2703135 -.2025521
noise | -.3447687 .0726024 -4.75 0.000 -.4870668 -.2024705
envir | -.348638 .0763919 -4.56 0.000 -.4983634 -.1989127
safe | -.1966342 .0272342 -7.22 0.000 -.2500122 -.1432562
holiday | -.6910101 .0683748 -10.11 0.000 -.8250223 -.556998
meal | -.1411564 .0402458 -3.51 0.000 -.2200368 -.0622761
expense | -.6599414 .0678141 -9.73 0.000 -.7928546 -.5270282
part | .0844039 .032096 2.63 0.009 .0214969 .147311
full | -.2935566 .0634737 -4.62 0.000 -.4179627 -.1691505
_cons | -4.303061 .1625748 -26.47 0.000 -4.621702 -3.984421
-------------+----------------------------------------------------------------
4 |
age | .0467059 .0079949 5.84 0.000 .0310362 .0623755
male | -.2364328 .0172864 -13.68 0.000 -.2703135 -.2025521
noise | -.7417132 .1929606 -3.84 0.000 -1.119909 -.3635173
envir | .002535 .2275338 0.01 0.991 -.4434232 .4484931
safe | -.1966342 .0272342 -7.22 0.000 -.2500122 -.1432562
holiday | -.824671 .1874773 -4.40 0.000 -1.19212 -.4572222
meal | -.1411564 .0402458 -3.51 0.000 -.2200368 -.0622761
expense | -.4722326 .1881537 -2.51 0.012 -.8410072 -.1034581
part | .0844039 .032096 2.63 0.009 .0214969 .147311
full | -.5355167 .1703173 -3.14 0.002 -.8693324 -.201701
_cons | -5.933492 .4521116 -13.12 0.000 -6.819614 -5.047369
------------------------------------------------------------------------------

I am using Stata 14.2 and gologit2 *! version 3.1.1 18oct2016 Richard Williams, [email protected].

If anyone knows/has some more output interpretation examples, it would be of great help.

Thank you in advance for any advice!

Vlad

[Richard Williams]

Welcome to Statalist. Note that your output is very hard to read. You should use code tags. See pt. 12 of the FAQ for how to post Qs effectively.

I suggest reading the following and then writing back if Qs persist:

https://www.dropbox.com/s/arkevwxlyf...t2016.pdf?dl=0 (Journal article on "Understanding and interpreting generalized ordered logit models")

http://www.stata-journal.com/article...article=st0097 (Stata Journal article that introduced the gologit2 command)

https://www3.nd.edu/~rwilliam/xsoc73994/Margins05.pdf (Handout that explains how to use margins and the spost13 commands to make results from ologit, mlogit, gologit2, and other commands more interpretable).

As far as signs being wrong we really can't comment because you haven't said how health is coded. If 1 = great health and 5 = poor health, the signs of effects may be perfectly reasonable. e.g. the positive coefficient for age would imply that health gets worse as you get older; the negative coefficient for expense would mean that those who can cover unexpected expenses tend to be in better health than those who can't. That all seems perfectly reasonable to me IF health is coded the way I just said.

Also, if you think these results are implausible, are the results from ologit more plausible? My guess is they aren't. But again, we don't have enough information to offer fully informed opinions here.