handling the false discovery rate ("fdr")

[16svale]

Dear MicrobiomeStat developer,
I have a question regarding the false discovery rate (FDR) correction.

I see on the documentation that the linda() function already takes into account FDR. However, in the function generate_taxa_volcano_single() there is an option called feature.mt.method = "none" or "fdr".

"feature.mt.method: There are two options available for this parameter: "fdr" (False Discovery Rate) and "none". Regardless of how this parameter is set, it's crucial to note that the generate_taxa_test_single function always performs adjustments post-testing. However, the feature.mt.method specifically influences the visualization in the volcano plot"

I am wondering what is the reason of adding another fdr layer on top of the LinDa calculation.
Could you provide an explanation on this?

Best regards

[cafferychen777]

Dear @16svale,

Thank you for reaching out with your question regarding the handling of the False Discovery Rate (FDR) correction in MicrobiomeStat.

Indeed, for high-dimensional data, incorporating FDR adjustment is essential to control for multiple testing issues. Regardless of the setting, the linda() function inherently conducts FDR adjustments. This adjustment is reflected in both the raw p-values and the adjusted p-values.

Regarding the feature.mt.method parameter, its purpose is to provide users with the option to choose between using adjusted p-values (via FDR correction) or raw p-values for visualization in the volcano plot generated by the generate_taxa_volcano_single() function. Opting for raw p-values in visualization is primarily for aesthetic reasons, ensuring the clarity and interpretability of the plot.

In summary, while FDR adjustment is integral to the statistical procedures performed by MicrobiomeStat, the feature.mt.method parameter offers flexibility in visualization choices without compromising the underlying statistical rigor.

Best regards,

Chen YANG