Overwhelming efficacy - the stats required, page-276

I think what he meant was that Bayesian theory is predictive and will apply why has been observed in the first 90 patients and predict how the rest of the patient population (even if not recruited) will play out.

so it’s not simple at all., because predictive models are highly sensitive to the inputs you add to the model, each basically teaching the model to be more accurate... it’s like the model learns from past experience and becomes more and more powerful as more data is input.

its also highly variable as you can see... the outcome of the analysis depends on what inputs are put in and modelled out.

this method of interim analysis is actually favourable to our trial IMO... if it really does work- this interim read out has a good chance of predicting that outcome.