911799_en-US

liliconcom

Exact Definition of Moving Box Plot

I really like to use the Moving Box Plot portion of Graph Builder.  I can't actually find in the documentation what the exact boundaries represent.   Based on a little trial and error they appear to line up with the local 5th, 25th, 50th, 75th, & 95th percentile, but I haven't been able to confirm that anywhere.   Can you please tell me what exactly these boundaries represent? or how they are calculated?

Data Exploration and VisualizationRe: Exact Definition of Moving Box Plot

This has been also asked here with no answer Moving Box Smoother (Graph Builder) . @XanGregg has this been documented somewhere? I haven't seen it mentioned in JMP Help or in New in Graph Builder for JMP 19. 

 

This is the only mention I quickly found from JMP Help

Moving Box

Indicates a method that smooths data by averaging values within a fixed-size window that moves across the data.

Re: Exact Definition of Moving Box Plot

Thanks for tagging me @jthi. I don't see any other docs right now either. Until that's sorted:

Moving Box is an generalization of John Tukey's "wandering schematic plot" where "schematic plot" was a name he used for the box plot. 

As I recall (I don't have his book handy), his version used 10 or so overlapping intervals and made a box plot for each interval and then connected the key values with lines.

Our generalization uses the local kernel smoother to compute moving versions of the 5 key values (lower whisker, lower quartile, median, upper quartile, and upper whisker). Except a whisker is hard to translate to a local kernel fitter, so we use 0.005 and 0.995 quantiles instead, which are about what a 1.5 IQR whisker would translate to for a Gaussian distribution.

Re: Exact Definition of Moving Box Plot