Hi, I think you can read General Information Biological Assay Validation, USP 40 (2017), in this ,1033> all re explained in detail, in addition you can read also Analytical Data Interpretation, USP 40, kind regards

First comes a problem with standard error. There are two meanings for SE. The first is simply the SD divided by square root of the sample size. The alternative other use is in regression where the SEM is the least squares estimate of the SD.

If by "mean+SEM" you are using the first definition, then it makes little difference so long as the sample size is clear.

If you are using the second definition then "mean+SEM" is a major mistake unless you intended to write "LSMean+SEM," where the LSMean is the least squares estimate of the mean.

The two acceptable choices are "mean+SD" and "LSMean+SEM" if one is using the second definition for SEM. In this case I prefer "mean+SD" because it tells me more about the raw data.

Consider: I run an experiment on grain yield given three factors: nitrogen, potassium, phosphorous. I provide you with a LSMean+SEM. In this case, the LSMean+SEM for yield have been corrected for the effect of potassium, phosphorous, and interactions. My friend Bob was clever and ran a similar experiment using potassium, phosphorous, and sulfur. He too gets an LSMean+SEM, but this has been corrected for different factors. While both values have units of kilograms per hectare, they are not easily comparable because they are adjusted by different factors and one does not know the effect of the missing factors in each model. Worse yet, both models have a long list of other factors that are missing: everything from irrigation to physicochemical soil properties. Everything is more complex if we used different transformations, or different methods (parametric versus nonparametric, OLS versus ML).

The very worst choice is to try and compromise: I used a log transformation so I will take my LSMeans and back-transform them. This approach yields grossly misleading numbers.