topic How is Std Error for Beta Binomial Distribution Analysis calculated? - differences between R and JMP in Discussions

How is Std Error for Beta Binomial Distribution Analysis calculated? - differences between R and JMP

M_Kraus — Fri, 09 Jun 2023 00:33:48 GMT

Hi All!

We are calculating beta binomial distribution parameters in R and wanted to compare the results and the std error.

Typically, the values for location and dispersion are up to 1% higher in JMP and there is also some discrepancy in the std error.

We would like to match here our R-code.

May anybody provide a hint how the standard error for mu and rho are calculated? Fisher Matrix?

Thanks

Mark_Bailey — Tue, 18 May 2021 13:48:54 GMT

Have you seen the JMP Help about this topic? Is it possible that the R function or model uses a different parameterization?

M_Kraus — Tue, 18 May 2021 14:01:50 GMT

I found this in your documentation:

Var(x) = np(1-p)[1+(n-1)δ]

Is this the Std Error shown in the table?

Mark_Bailey — Wed, 19 May 2021 14:04:52 GMT

No, that quantity is the variance of the random variable when assuming a beta binomial distribution.

tonya_mauldin — Wed, 19 May 2021 17:36:03 GMT

JMP uses the inverse of the information matrix to estimate the covariance matrix (and standard errors). https://en.wikipedia.org/wiki/Fisher_information

M_Kraus — Thu, 20 May 2021 07:08:38 GMT

Thank you very much. This is what I thought.