Bayesian Process Capability

[vandenman]

No description provided.

[hrgodmann]

I'll split between Code and JASP comments.
Code (minor) Comments:

• In the R code, in function .bpcsSamplePosteriorOrPrior , is the MCMC settings supposed to be hardcoded?
• In Description.qml, should it be Gauge r&R Study really or actually Gauge R&R Study?
• In the priors qml, should the console.log functions stay in?

JASP comments:

• specifying a higher "lower specification limit" than the "higher specification limit" just transforms the lower into a negative number, I assume this is not desired behaviour.
• When I run the analysis, then change lower and upper specification limit, the Capability table doesn't update, also the plots "Posterior Distribution" dont update.
• When in the "Prior and Posterior" inference section the "Posterior distribution" plot is chose, and then the Layout 'All metrics in one plot" is selected, additionally printing "point estimates" and "CI" in the plot makes it unreadable even when widened because the values seem randomly distributed across the plot.
• I think plotting the priors looks a bit strange because the x axis is so wide, thus the distribution is not really identifiable by eye anymore. this means that the central CI plotting function for the prior basically does nothing for Cp CpU and CpL. If "identical across pane" is chosen for the axes, it makes it even worse
• sequential plots are awesome!!
• do the MCMC settings do anything yet, as they are still hard coded in the code I believe
• the "posterior predictive distribution" plot shows wide samples as a histogram but very narrow CI when chosen, is this correct?

[julianwuth]

I spotted some things when working with the sequential analysis:

1. The "Interval estimate plot" is missing the "Lower" and "Upper" fields as dependencies.
2. The automatically generated y-limits for the "Interval estimate plot" should just be 0 and 1 I assume.