Suggestions

[PaulSoderlind]

Just noticed your package. Looks promising. Just two suggestions

1. drop Optim.jl since its a heavy dependency. I believe a simple golden section search should be enough (like 20 lines of code). See how KernelDensity.jl did this.
2. add standard errors, eg. as in my teaching notebook

[tbeason]

Thanks! I just whipped this up as there was nothing that existed and I needed the functionality. It passes my tests but has not been thoroughly tested against R/python equivalents.

1. I agree. I have at times considered making a small "OptimLite" package that contains just the univariate optimization functionality, precisely for this reason.
2. Yes I had started to look into standard errors. It seems there are several ways to do it. Perhaps I should focus on the non-bootstrap approach like you have in your notes first as that doesn't really require anything extra. Hall & Horowitz (2013) seemed to suggest that it was biased however.

Open to more suggestions and PRs if you find the package useful!