Experiments with online bayesian calculators

I’ve been doing experiments with Bayesian calculators online. Since this is not very common I wanted to ask you for some comments / feedback. The prototype is provisionally hosted at:


It takes a few seconds to reconstruct the Bayesian model.
Maybe I’ll change the plots to make them clearer.
I wanted to ask you what you think about the general idea of our experiment.

A lognormal AFT model was used. The prior of a previous meta-analysis was used for the therapeutic effect. Three interactions have been applied: treatment + treatment * spline (year) + treatment * spline (age) + treatment * spline (histology) + covariates. The database has >1000 events, and seems to support this.

The calculator basically measures the relative effect (time ratio) of two strategies used in gastric cancer. The model incorporates the variable year of treatment, therefore, the objective is not predictive but explanatory, to check the external validity of RCTs that have been published. The reason for this is that in 2006, one of the strategies appeared to be far superior, but over time this effect seems to have diminished. We wondered why, and whether there were any patients who currently still benefited.

1 Like

There was another topic very similar to this one. You may want to merge this with that.

1 Like

The truth is, I’m having trouble publishing articles based on Bayesian analysis. I don’t think the reason is the low quality of the manuscripts, which are similar to the frequentist articles that we do publish quite easily. The reason is that the reviewer I think is not familiar with the method, and is afraid to accept a publication that might be wrong but cannot evaluate. Therefore he prefers to reject it even if it is a meritorious contribution, rather than taking responsibility for accepting a wrong article. I perceive this by reading between the lines of the reviews. The reasons for rejection are sometimes pathetic.

There is a publication bias against articles based on Bayesian analysis. Therefore, coauthors often get nervous about this.