The MATLAB function in TSK.m analyzes a dose-response curve and calculates the median dose (for instance, the lethal dose for 50% of the population (LD50)) using the Trimmed Spearman-Karber method. This is an alternative to dose-response analysis methods such as probit and logit methods. It does not require glmfit or the stats toolbox.
All calculations in this code are based on the method from Hamilton, M.A.; Russo, R.C.; Thurston, R.V. Trimmed Spearman-Karber Method for Estimating Median Lethal Concentrations in Toxicity Bioassays. Enviro. Sci. Tech. 1977, 11 (7), 714-719: http://dx.doi.org/10.1021/es60130a004. See also ibid, 1978, 12 (4), 417: http://dx.doi.org/10.1021/es60140a017.
This code was validated using the data given in Hamilton et al. (1977) and the example given in the documentation for the EPA's DOS program to perform the Spearman-Karber method. (Note that the confidence intervals given in Hamilton et al. (1977) are incorrect, as per Hamilton et al. (1978).) This code replicates the EPA program's results. Hamilton's test cases are provided in TSKTest.m.
Someone asked me via e-mail if it would be better to use the Student's distribution instead of the normal distribution to calculate the confidence intervals. This isn't what was done in Hamilton, but if you want to modify the code and have an install of the stats toolbox, change line 248 to:
v = tinv((1-conf)/2,s+1);
Note: I don't have a copy of the stats toolbox and have not tested this.
adding BSD organization clause
text editing, adding URLs: no changes to code