Katie Harding <katiemharding@...> writes: > > I have insect size measurements that I fit to a Weibull curve > to determine EC50. This is my model > fit:model=nls(y~SSweibull(x,asym,drop,lrc,pwr)) > I need to use the untreated buffer samples as the maximum size > (the left most asymptope). If I put the doses > as 0,1,2,4,8,16,32,64,128 ppm, > I get error messages indicating that 0 values for x are inappropriate. > This is the error message:Error in qr.default(.swts * > attr(rhs, "gradient")) : NA/NaN/Inf in > foreign function call (arg 1) > I found that I get more consistent results when I put > 0.01 for the dose of the untreated controls. I was > wondering if there was a convention to follow when using > untreated buffer samples in EC50 calculations. Cheers,Katie
Are you measuring a mortality response? If so, you should probably be allowing the responses to be binomial rather than using least-squares (as nls() does). The drc package from CRAN might be helpful (it has a built-in weibull response function). Can you be slightly more specific/give a reproducible example? _______________________________________________ R-sig-ecology mailing list R-sig-ecology@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology