Yes, it is possible to reflect multiple correlation coefficients (e.g., Spearman's, Pearson's, point-biserial, and phi) in a single correlation heatmap, but it requires careful design to avoid confusion. Since you’ve used different correlation coefficients based on the nature of your variables (BMI, diabetes mellitus, arterial hypertension, blood glucose, and total cholesterol levels), you can create a heatmap that either combines them into a single visualization with clear labeling or presents them in separate panels. With R:

library(ggplot2)

library(reshape2)

# Example dataset (replace with your actual data)

set.seed(123)

data

It’s generally not a good idea to combine different types of correlation coefficients in one single heatmap. The main reason is that each of these coefficients is meant for different types of variable pairs and they aren't really comparable in a meaningful way.

For example, Pearson is for continuous variables with linear relationships, Spearman works better for monotonic relationships (and doesn't assume normality), point-biserial is for binary vs. continuous, and phi is for binary vs. binary. If you throw all of them into one heatmap, it might look visually clean, but it can easily mislead readers because it suggests that all correlations are calculated the same way, when they’re not.

What I’d suggest instead is either:

• Creating separate heatmaps depending on the variable types (e.g., one for continuous-continuous, one for binary-continuous, etc.), or
• Creating a matrix or table where you list the correlation values along with the type of coefficient used for each pair, you can even annotate a heatmap this way if you really want everything in one visual.

This way, you’re being transparent about the methods and avoiding confusion. It’s also more helpful for readers who might want to interpret or replicate your analysis.

It's unusual, but I don't see any reason why you can't.

Those correlation coefficients are on the same scale (-1, 0, +1).

They may not be directly comparable, though. I mean, is a Spearman's rho of 0.6 stronger than a Pearson's r of 0.5 ? I probably wouldn't consider things that finely. But other than that, I think it all makes sense.

I suggest you to consult the following paper:

S. Strings, C. Wells, C. Bell, A.J. Tomiyama (2023).

The association of body mass index and odds of type 2 diabetes mellitus varies by race/ethnicity, Public Health,

Volume 215, Pages 27-30,

https://doi.org/10.1016/j.puhe.2022.11.017.