Simple scripts to illustrate challenges of inference with noisy data
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
.gitignore
P-values-0.005.pdf
P-values-0.005.png
P-values.R
P-values.pdf
P-values.png
Readme.md
bayesian-intervals.txt
bayesian-tight-priors.txt
bayesian.pdf
bayesian.png
credible-intervals.R
point.stan
tight-priors.png

Readme.md

Noisy data

P-values.R illustrates how likely it is that significant P-values are misleading when the magnitude of a between group difference is small relative to within group variation. P-values.pdf was produced by running this code.

credible-intervals.R illustrates how a Bayesian estimate of the difference between the mean of (replicates of) the same populations behaves with respect to coverage (the fraction of replicates in which the 95% credible intervals contain the true difference) and sign (the fraction of time in which the estimate of the difference has the same sign as the true difference). The Bayesian model is implemented in Stan (point.stan).