topic How to calculate REML Estimates of Varaince Componenets based on USP <1033> Biological Assay Validation? in Discussions

How to calculate REML Estimates of Varaince Componenets based on USP <1033> Biological Assay Validation?

Luk_s — Wed, 16 Oct 2024 09:40:52 GMT

Hello,

I have encountered a problem with calculating the results from REML analysis based on the data and results presented in USP <1033>. I have trouble with computing the data in Table. 4 and getting the same results contained in Table 9. Is there a person who can handle this data and explain how to proceed in making the REML analysis using this example?

Thanks!

Re: How to calculate REML Estimates of Varaince Componenets based on USP <1033> Biological Assay Validation?

Luk_s — Wed, 16 Oct 2024 09:45:03 GMT

For more information, I provide the sources: 1033 BIOLOGICAL ASSAY VALIDATION

Re: How to calculate REML Estimates of Varaince Componenets based on USP <1033> Biological Assay Validation?

MRB3855 — Wed, 16 Oct 2024 12:14:49 GMT

Hi @Luk_s , and welcome. I'm able to reproduce tables 5 and 6. But, not 9. It appears the results in table 9 are an example using different data. That can be seen by the difference in Var(error). In table 5 it is 0.000766. In table 9 it is 0.0022. They should be the same if both using the same data. So, table 9 is not based on the data shown in table 4. It is an example from something else, as implied by the statement " Variance component estimates obtained from such an analysis are presented in Table 9" (italics mine); notice it does not refer back to the data in table 4. And, as they correctly point under table 5, " Analyst and media lot have not been included because of the small number of levels (2 levels) for each factor." , so it would be ill-advised to carry out the type of analysis shown in table 9.

FWIW, and as ill-advised as this is, I get the following (Table 4, Level 0.5):

Re: How to calculate REML Estimates of Varaince Componenets based on USP <1033> Biological Assay Validation?

Luk_s — Wed, 16 Oct 2024 12:46:08 GMT

Hi @MRB3855 thank you kindly for your attention to the problem which I have presented. I was convinced that I was doing something wrong in developing the data for the model. I also did a check and we come out with the same results, which I'm really pleased about. I have dealt with the previous data table ones as well, however, I was surprised by the results obtained in this example.

Thanks!