One way to do statistics less wrong is to move from statistical testing to statistical modelling. This is what we are trying to support with BayesHive at https:

One way to do statistics less wrong is to move from statistical testing to statistical modelling. This is what we are trying to support with BayesHive at https://bayeshive.com https://bayeshive.com Other ways of doing this include JAGS (http://mcmc-jags.sourceforge.net/ http://mcmc-jags.sourceforge.net/) and Stan (http://mc-stan.org/ http://mc-stan.org/) The advantage of statistical modelling is that it makes your assumptions very explicit, and there is more of an emphasis on effect size estimation and less on reaching arbitrary significance thresholds.