Hey, thank you for this tool. It is very convenient to use
Recently I redraw a heatmap from Maaslin2 output in python and used a formula as in the code to get values:
−log(qval)*sign(coeff)
I like very much the implementation of this visualization.
I am not a statistician and wondering if I can use the same formula in other analysis.
For example, I performed multiple spearman correlations on some data, and got a list of p-values. Can I adjust p-values by Benjamini & Hochberg procedure to get q-values, and then apply the same formula and create a heatmap, or this formula is specific for the analysis in Maaslin2?
I already run it and got values in range from -13 to +13. Or it is not advisable to do it like this? 
Hello, thanks for your question. The visualization you describe would not be inappropriate, per se, but it has the disadvantage that the effect size (i.e. the value of the correlation coefficient) is not included. Usually we visualize the results from Spearman correlations using a heatmap with color corresponding to the strength and direction (sign) of the correlation, with an indicator such as an asterisk to denote statistically significant boxes (or increasing numbers of asterisks to indicate progressively smaller p-values) in the heatmap. That way, both the statistical significances and the effect sizes of the associations are evident from the plot.
1 Like
Hi, thank you for a nice explanation! Now it is more clear to me how to visualize it properly 