I have a three-part question from a logistic regression. In the case of this study, I have three categorical predictors: gender (M/F), age group (2, 3, 4), and country of origin (1,2,3,4,5). Both the logit and odds ratio outputs are below.
Here are my questions:
1.) Which confidence intervals do I report? The ones associated with the coefficient or with the odds ratio?
2.) What happens when CI and p-values don't agree? For example, in the first output, ageGroup 3 has an OR = 0.207-1.2, which crosses the "1" threshold, but has a p=0.005. Crossing the 1 means insignificant, correct? But p<0.05 is significant. Does one trump the other? Or does this mean something is wrong with my data?
3.) This might be answered by #2, but how do I interpret an OR greater than 2? I know and OR of 0.76 can be interpreted at "24% less likely" (correct?), but what about and OR of 2.02?
Thanks
Code:
logit stage4 i.ageGroup i.country i.sex Iteration 0: log likelihood = -612.81892 Iteration 1: log likelihood = -580.34846 Iteration 2: log likelihood = -579.85354 Iteration 3: log likelihood = -579.85215 Iteration 4: log likelihood = -579.85215 Logistic regression Number of obs = 940 LR chi2(7) = 65.93 Prob > chi2 = 0.0000 Log likelihood = -579.85215 Pseudo R2 = 0.0538 ------------------------------------------------------------------------------ stage4 | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- ageGroup | 3 | .7038665 .2534184 2.78 0.005 .2071756 1.200557 4 | 1.47736 .253085 5.84 0.000 .981322 1.973397 | country | 2 | -.2736965 .208729 -1.31 0.190 -.6827979 .135405 3 | -.2919848 .2617669 -1.12 0.265 -.8050386 .2210689 4 | -.1398201 .2613874 -0.53 0.593 -.6521299 .3724897 5 | -.7572049 .354703 -2.13 0.033 -1.45241 -.0619997 | 1.sex | -.3730072 .1425399 -2.62 0.009 -.6523802 -.0936342 _cons | -1.228277 .24262 -5.06 0.000 -1.703803 -.7527503 ------------------------------------------------------------------------------
Code:
logit stage4 i.ageGroup i.country i.sex, or Iteration 0: log likelihood = -612.81892 Iteration 1: log likelihood = -580.34846 Iteration 2: log likelihood = -579.85354 Iteration 3: log likelihood = -579.85215 Iteration 4: log likelihood = -579.85215 Logistic regression Number of obs = 940 LR chi2(7) = 65.93 Prob > chi2 = 0.0000 Log likelihood = -579.85215 Pseudo R2 = 0.0538 ------------------------------------------------------------------------------ stage4 | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- ageGroup | 3 | 2.021554 .512299 2.78 0.005 1.230199 3.321968 4 | 4.381361 1.108857 5.84 0.000 2.667981 7.195077 | country | 2 | .7605629 .1587516 -1.31 0.190 .5052015 1.145 3 | .7467799 .1954823 -1.12 0.265 .4470707 1.247409 4 | .8695146 .2272801 -0.53 0.593 .520935 1.451344 5 | .4689754 .166347 -2.13 0.033 .2340056 .9398832 | 1.sex | .6886603 .0981616 -2.62 0.009 .5208047 .9106158 _cons | .2927967 .0710383 -5.06 0.000 .1819901 .4710692 ------------------------------------------------------------------------------ Note: _cons estimates baseline odds.
0 Response to Odds ratio and Con
Post a Comment