-heatplot- update: -fast- option for big data

[Ben Jann]

An update to the heatplot package is now available from SSC. To install the update, type

. ssc install heatplot, replace
. ssc install palettes, replace
. ssc install colrspace, replace

heatplot now has a fast option that can be used to speed up computations in large datasets. If fast is specified, some of the computations are performed by commands from the gtools package by Mauricio Caceres Bravo. To install gtools, type

. ssc install gtools, replace
. gtools, upgrade

Here are some example timings (in seconds) from my computer for a dataset with 5 million observations; I specify option nograph so that the timings only include the computations but not the drawing of the graph.

. set rmsg on
r; t=0.00 8:41:27

. set obs 5000000
number of observations (_N) was 0, now 5,000,000
r; t=0.06 8:41:27

. drawnorm y x
r; t=0.92 8:41:28

. heatplot y x, nograph
r; t=14.18 8:41:42

. heatplot y x, nograph fast
r; t=2.52 8:41:45

. generate byte g = ceil(runiform()*20)
r; t=0.15 8:41:45

. heatplot y i.g, nograph
r; t=21.09 8:42:06

. heatplot y i.g, nograph fast
r; t=2.66 8:42:09

ben

[eric_a_booth]

Ben Jann -heatplot- (especially the hexagon plotting!) is fantastic. Thank you!

[Ben Jann]

Thanks!
ben