CI for dependent variable values - proc logistic

acardenz — Thu, 06 May 2021 16:45:45 GMT

Hello,

I have data with a binary response (event, no event) and a continuous dependent variable (e.g. dose).

I am interested in the value needed of the dependent variable to achive a probability of response of 30%.

I have used the model estimates from proc logistic to identify the point esimtate for the dependent variable (e.g. with my data a dose of 252mg produces a predicted probability of event of 30%). In addition to the point estimate, I am interested in obtaining a 95% confidence interval for this value. Lower= and upper= options in the output statement give the range for the probability at a given dependent variable value. I want the opposite - I want the range for the dependent variable at a given probability. Is that possible?

Re: CI for dependent variable values - proc logistic

Rick_SAS — Thu, 06 May 2021 17:09:40 GMT

I believe this is known as the LD30 value (lethal dose=0.3), You can use PROC PROBIT to obtain a point estimate for the X value as well as an (inverse) CI for the value. See the last example at

https://blogs.sas.com/content/iml/2020/12/07/transplant-indoor-christmas-tree.html

which includes a link to the PROC PROBIT documentation. Note that the example uses LD50, so you'll have to modify the INVERSECL option to be

INVERSECL(prob=0.3)

topic CI for dependent variable values - proc logistic in Statistical Procedures

CI for dependent variable values - proc logistic

Re: CI for dependent variable values - proc logistic