search

nhs-r-community / NHSRplotthedots

An SPC package to support NHSE/I 'Making Data Count' programme
https://nhs-r-community.github.io/NHSRplotthedots/
Other
48 stars 23 forks source link

Add other methods? #127

Open ChrisBeeley opened 3 years ago

ChrisBeeley commented 3 years ago

Mohammed asked

I would like to suggest incorporating Jacob's run chart into the plot the dots package

Thoughts?

tomjemmett commented 3 years ago

one thing I think we should be careful of is to try to follow the Unix philosophy (do one thing, do it right). This is designed to do the NHS Making Data Count stuff, so we probably shouldn't go to far into other territories.

I think that sits more in with the stuff that I wanted to build before but haven't got round to yet (building a very generic spc package that does calculations and nothing much else).

ThomUK commented 2 years ago

I would like to see run charts within this package, and could make use of them now. Ideally with a slightly different visual appearance (perhaps a background colour by default) to separate it in people's minds from the SPC outputs when used in the same reports.

Rare-events are the other chart I need, but I'm going to plot XmR charts of time-between-events rather than use a dedicated T or G chart for the time being.

ThomUK commented 2 years ago

There's been a new request on slack, with the pertinent bit copied below:

However I would really like if you could change the chart type, for example to a P or C chart instead of an XMR chart. I guess this would just be a case of changing how the control limits are calculated. Is this something I could do myself withe the code that makes the function or is something you guys can help with?

My own view on this is that we need to be careful broadening our scope beyond what the Making Data Count team teach. This package was originally intended as a companion to the MDC programme, in keeping with its teaching and recommendations. In particular MDC follow some of the pragmatic advice of Don Wheeler, who refers to XmR as a swiss army knife, because it makes no assumptions about the distribution of the underlying data (binomial for the p chart, and poisson for the c chart), and can be used in place of several of the more specialised charts.

There is an article by Wheeler on this called "what about the p chart?" Two links below (unfortunately both requiring signup): https://future.nhs.uk/MDC/view?objectId=105047429 https://www.qualitydigest.com/inside/quality-insider-article/what-about-p-charts-093011.html

So my view is that implementing p or c charts would take us away from the MDC programme in a way that may not be helpful. The MDC visual style and teaching is designed for interpretation often by non-statistical users, who rely on the visual consistency of the output to "trust" what they are seeing.

Run charts would have the same problem, and even though I would like to implement them somewhere, if we did in this package we'd need to be careful that the output looked different enough that there could be no confusion. (how to avoid questions like - Where have the process limits gone? or - Why are the process limits not straight lines any more?)

LynHoward commented 2 years ago

MDC include P, T & G charts as well as XmR, so I don't think adding these would be outside the scope of a MDC aligned R package. I'd really like to see P charts implemented.

For run charts, perhaps a badge to indicate the absence of SPC limits? There are probably accessibility issues with having a background colour having meaning.

ChrisNyunt commented 11 months ago

MDC include P, T & G charts as well as XmR, so I don't think adding these would be outside the scope of a MDC aligned R package. I'd really like to see P charts implemented.

For run charts, perhaps a badge to indicate the absence of SPC limits? There are probably accessibility issues with having a background colour having meaning.

Same here, it would be very useful to have P charts implemented in the package.