Good morning, is there any way to plot nomograms from brms-fitted models?

For example, would it be correct to predict the linear predictor of a brms model using ordinary least square regression to draw nomograms with the rms package if R2=1? Any other option?

Thanks.

Hi,

first I don’t think there is anything ready-made for this. Unfortunately, I am not sure what you mean by

Could you elaborate? Do you have an example of a nomogram that works similarly as you envision?

However, as far as I understand, nomograms (at least the few ones I’ve seen) cannot easily show uncertainty, but uncertainty is an important part of the brms fit (or any Bayesian modelling). So I believe you will need to get creative. If you don’t want to plot the uncertainty, you can directly use the MAP estimate (via `optimizing`

) or just use a frequentist package…

I am really not sure who might have some nomogram-related expertise around here, maybe @sakrejda? It is really a fun question to ponder (at least to me), but I don’t see any simple solutions.

I’ve also changed the category of the post as it (very distantly) relates to brms .

I’ve been thinking about this, and maybe the best thing is not to make the nomogram, bet develop a web basel prognostic calculator based on the underlying brms model. Thus I could use the brms.fitted function to calculate the uncertainty of predictions through a posterior distributions. However, if someone had done it it would be interesting to guide me. Thanks.