Issues interpreting coefplot after generating a partial proportional odds ordinal logistic model.
Hi everyone,
I would just like to clarify about the use of coefplot after generating a Partial Proportional Odds Ordinal Logistic model using gologit2.
I generated a partial proportional odds model using gologit2 after finding that my model for outcome Q12 violates the proportional odds assumption. The model computes odds ratios for each experimental arm with arm 8 as the reference.
The following is my code:
gologit2 Q12 Q3A ib8.Arm, or
My output shows that some of the odds ratio for the experimental arms (e.g. arm 4 & 13) are different for each cut off point(I have 3 cut off points in the dependent variable). I have omitted the output for Q3A and _cons in the output below.
However, when I executed the coefplot command:
coefplot, eform drop(Q3A _cons) xline(1)
...it generates the following odds ratio plots, with only a single point estimate (below).
My question is, how do I interpret these single point estimates, when my partial proportional odds models have various odds ratios at different cut off points? Is the single point estimate an average odds ratio of a particular arm? I have been trying to find the answer online but failed. Any expert guidance on this matter is much appreciated. Thank you very much!
I would just like to clarify about the use of coefplot after generating a Partial Proportional Odds Ordinal Logistic model using gologit2.
I generated a partial proportional odds model using gologit2 after finding that my model for outcome Q12 violates the proportional odds assumption. The model computes odds ratios for each experimental arm with arm 8 as the reference.
The following is my code:
gologit2 Q12 Q3A ib8.Arm, or
My output shows that some of the odds ratio for the experimental arms (e.g. arm 4 & 13) are different for each cut off point(I have 3 cut off points in the dependent variable). I have omitted the output for Q3A and _cons in the output below.
| Q12 | exp(b) | Std. Err. | z | P>z | [95% Conf. | Interval] |
| Definitely No | ||||||
| Arm | ||||||
| 1 | .6852644 | .1560102 | -1.66 | 0.097 | .4386013 | 1.070647 |
| 2 | 1.100562 | .2524957 | 0.42 | 0.676 | .7019856 | 1.725444 |
| 3 | .9526014 | .2201017 | -0.21 | 0.834 | .605673 | 1.49825 |
| 4 | .2128189 | .089902 | -3.66 | 0.000 | .0929894 | .4870651 |
| 5 | 1.102713 | .2497955 | 0.43 | 0.666 | .7073617 | 1.71903 |
| 6 | .9582241 | .2176879 | -0.19 | 0.851 | .6138926 | 1.495691 |
| 7 | .865716 | .1986142 | -0.63 | 0.530 | .5521934 | 1.35725 |
| 9 | .8472162 | .1917797 | -0.73 | 0.464 | .5436413 | 1.32031 |
| 10 | .8490981 | .1937738 | -0.72 | 0.474 | .5428803 | 1.328041 |
| 11 | .5910764 | .135689 | -2.29 | 0.022 | .3769117 | .9269315 |
| 12 | .7848461 | .1785581 | -1.06 | 0.287 | .5024932 | 1.225854 |
| 13 | .3507189 | .1593813 | -2.31 | 0.021 | .1439255 | .8546352 |
| 14 | .7711762 | .1788714 | -1.12 | 0.263 | .4894638 | 1.215029 |
| Maybe no | ||||||
| Arm | ||||||
| 1 | .6852644 | .1560102 | -1.66 | 0.097 | .4386013 | 1.070647 |
| 2 | 1.100562 | .2524957 | 0.42 | 0.676 | .7019856 | 1.725444 |
| 3 | .9526014 | .2201017 | -0.21 | 0.834 | .605673 | 1.49825 |
| 4 | .5377783 | .1854247 | -1.80 | 0.072 | .2735967 | 1.05705 |
| 5 | 1.102713 | .2497955 | 0.43 | 0.666 | .7073617 | 1.71903 |
| 6 | .9582241 | .2176879 | -0.19 | 0.851 | .6138926 | 1.495691 |
| 7 | .865716 | .1986142 | -0.63 | 0.530 | .5521934 | 1.35725 |
| 9 | .8472162 | .1917797 | -0.73 | 0.464 | .5436413 | 1.32031 |
| 10 | .8490981 | .1937738 | -0.72 | 0.474 | .5428803 | 1.328041 |
| 11 | .5910764 | .135689 | -2.29 | 0.022 | .3769117 | .9269315 |
| 12 | .7848461 | .1785581 | -1.06 | 0.287 | .5024932 | 1.225854 |
| 13 | 1.445648 | .5312158 | 1.00 | 0.316 | .7035275 | 2.970598 |
| 14 | .7711762 | .1788714 | -1.12 | 0.263 | .4894638 | 1.215029 |
| Maybe yes | ||||||
| Arm | ||||||
| 1 | .6852644 | .1560102 | -1.66 | 0.097 | .4386013 | 1.070647 |
| 2 | 1.100562 | .2524957 | 0.42 | 0.676 | .7019856 | 1.725444 |
| 3 | .9526014 | .2201017 | -0.21 | 0.834 | .605673 | 1.49825 |
| 4 | .7674514 | .1927372 | -1.05 | 0.292 | .4691166 | 1.255512 |
| 5 | 1.102713 | .2497955 | 0.43 | 0.666 | .7073617 | 1.71903 |
| 6 | .9582241 | .2176879 | -0.19 | 0.851 | .6138926 | 1.495691 |
| 7 | .865716 | .1986142 | -0.63 | 0.530 | .5521934 | 1.35725 |
| 9 | .8472162 | .1917797 | -0.73 | 0.464 | .5436413 | 1.32031 |
| 10 | .8490981 | .1937738 | -0.72 | 0.474 | .5428803 | 1.328041 |
| 11 | .5910764 | .135689 | -2.29 | 0.022 | .3769117 | .9269315 |
| 12 | .7848461 | .1785581 | -1.06 | 0.287 | .5024932 | 1.225854 |
| 13 | .984204 | .2462258 | -0.06 | 0.949 | .6027449 | 1.607077 |
| 14 | .7711762 | .1788714 | -1.12 | 0.263 | .4894638 | 1.215029 |
coefplot, eform drop(Q3A _cons) xline(1)
...it generates the following odds ratio plots, with only a single point estimate (below).
My question is, how do I interpret these single point estimates, when my partial proportional odds models have various odds ratios at different cut off points? Is the single point estimate an average odds ratio of a particular arm? I have been trying to find the answer online but failed. Any expert guidance on this matter is much appreciated. Thank you very much!
