why my CI 95% is coming .42 - 4.38 but SE is .81 ?

[Shobhit Piplani]

why my CI 95% is coming .42 - 4.38 but SE is .81 ?
------------------------------------------------------------------------------

| Linearized

transfusion | Odds ratio std. err. t P>|t| [95% conf. interval]

-------------+----------------------------------------------------------------

treat | 1.369022 .8122529 0.53 0.597 .4276635 4.382469

_cons | .0413461 .0057798 -22.79 0.000 .0314324 .0543865

------------------------------------------------------------------------------

[Jared Greathouse]

Please read the FAQ to learn how to ask a question

[Ken Chui]

Because odds ratio is in a log scale so the CI is not computed as b +/- 1.96*SE. It's easier to see that in logit output rather than in logistic output.

If you rerun this with logit, the SE for treat should be about 0.593, and the coefficient of treat should be about 0.314. Assuming you have a big sample, the lower and upper bounds should be around:

. display 0.314 - 1.96*0.593
-.84828

. display 0.314 + 1.96*0.593
1.47628

Now if you covert them back to OR, you'd see why 0.42 and 4.38:

. display exp(0.314 - 1.96*0.593)
.42815072

. display exp(0.314 + 1.96*0.593)
4.3766343