I am getting conflicting results for probplot with weibull and censors. Is there some pre-processing going on in the back-ground that I am unaware of. This is when compared to Minitab.

When I run: parmhat=wblfit(X(:,1),.05,X(:,2)) where X(:,1) are survival hours and X(:,2) is a censoring column, I get the same parameters that Minitab calculates. However, when I use: probplot('wbl',X(:,1),X(:,2)) Matlab rejects the null hypothesis that the data is weibull whereas Minitab finds it fits very well. I can't seem to understand why.

Answer 1

[Jeff Miller]

I don't think probplot does any fitting. If not, your probplot command would just plot your data against a Weibull with the default parameter values. If those values are not appropriate for your data, then probplot will tell you that Weibull doesn't fit.