I would like to estimate a confidence interval for a proportion when there are zero counts. I find that -ci prop- will give a CI estimate, while the -proportion- estimate command never does, and I wondered why. Perhaps I missed it but the manual suggests that both commands estimate the SE/confidence intervals in similar ways.
Code:
set obs 100 gen x= _n>50 gen y= _n >75 tab x y proportion y, over(x) citype(wilson) /* Proportion estimation Number of obs = 100 -------------------------------------------------------------- | Wilson | Proportion Std. Err. [95% Conf. Interval] -------------+------------------------------------------------ y@x | 0 0 | 1 0 . . 0 1 | .5 .0707107 .3664451 .6335549 1 0 | 0 (no observations) 1 1 | .5 .0707107 .3664451 .6335549 -------------------------------------------------------------- */ ci prop y if x==0, wilson /* ------ Wilson ------ Variable | Obs Proportion Std. Err. [95% Conf. Interval] -------------+--------------------------------------------------------------- y | 50 0 0 6.94e-18 .0713476 */ ci prop x if y==1, wilson /* ------ Wilson ------ Variable | Obs Proportion Std. Err. [95% Conf. Interval] -------------+--------------------------------------------------------------- x | 25 1 0 .8668077 1 */
Thanks,
Scott
0 Response to ci prop and proportions with zero counts
Post a Comment