The current textbook has been written as a help to medical / health
professionals and students for the study of modern Bayesian statistics,
where posterior and prior odds have been replaced with posterior and
prior likelihood distributions. Why may likelihood distributions better
than normal distributions estimate uncertainties of statistical test
results? Nobody knows for sure, and the use of likelihood distributions
instead of normal distributions for the purpose has only just begun, but
already everybody is trying and using them. SPSS statistical software
version 25 (2017) has started to provide a combined module entitled
Bayesian Statistics including almost all of the modern Bayesian tests
(Bayesian t-tests, analysis of variance (anova), linear regression,
crosstabs etc.).
Modern Bayesian statistics is based on biological likelihoods, and may
better fit clinical data than traditional tests based normal
distributions do. This is the first edition to systematically imply
modern Bayesian statistics in traditional clinical data analysis. This
edition also demonstrates that Markov Chain Monte Carlo procedures laid
out as Bayesian tests provide more robust correlation coefficients than
traditional tests do. It also shows that traditional path statistics are
both textually and conceptionally like Bayes theorems, and that
structural equations models computed from them are the basis of
multistep regressions, as used with causal Bayesian networks.
