Hi I am trying to get the starting values for non-linear regression of a data set that follows a weibull function. I tried to do the OLS regression of this equation:

log(log(1-F(t)))=-Beta(log(lambda))-Beta(log(t)) to get the starting values.

The only problem is that F(t) is greater than 1 which leads taking a log of a negative values (not possible.) Anyother suggestions on getting these starting values.

Look att Wikipedia on the Weibull distribution and write the equation like in the Weibull plot section. You can do the maximum likelihood estimation and use a starting value of say k=3.