I am trying to use a glm with a binomial distribution (logit link) to analyse data from a dose response curve for the lethal effects of different bacterial strains. I now want to obtain estimates of the lethal dose 50 (LD50).

My model is simply of the form:

survival ~ a + b*dose

Thus since I have a logit link, obtaining the estimates for the LD50 (i.e. p=0.5) simply means dividing -a/b. However, I am also wanting to know the confidence intervals for these estimates and that's where I am unsure about what to do. If anyone would have a suggestion for this, that would be great.

  • $\begingroup$ Depending on your sample size, you could use bootstrapping to obtain CIs. $\endgroup$
    – boscovich
    Nov 30, 2012 at 19:10
  • $\begingroup$ Hi Andrea, Thanks for your reply. I have 5 datapoints per dose (4 doses). Could you please further explain how the bootstrapping could be done in this case? (sorry, I haven't really used this much). Cheers, Ricardo $\endgroup$
    – ramiro
    Dec 1, 2012 at 14:37
  • $\begingroup$ Have a look at this link. $\endgroup$
    – crogg01
    Feb 2, 2014 at 3:03

1 Answer 1


There is a package "drc" for estimation of LD50 and its confidence intervals.

In another way, your glm results can be treated by a simple function "dose.p" in package MASS that calculates LD50 and its SE.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.