combining histograms of different bin sizes and plotting gamma distribution on them

[muhammad adeel]

Hi all

I have prepared histograms (attached) of % trip by trip lengths in the UK for two modes of travel: car and surface rail. Now the challenge is:

1. To generate trendline of Gamma distribution for each histogram.

2. To overlay both histograms/trendlines in one plot (note different bin sizes)


Any help or alternate method, please.

Thanks

Adeel

ITS Leeds university

[Scott Merryman]

1. I don't know what a trendline of Gamma distribution means.
2. You can use -twoway histogram- to overlay figures:

sysuse auto
twoway histogram mpg , w(5)  color(green%40) || histogram trunk, color(blue%30) w(2)

[Nick Cox]

Comparing a curve and a histogram are not nearly so effective as a dedicated gamma quantile plot. See qgamma on SSC.

[muhammad adeel]

Hi Scott, Thanks, histogram overlay works but transparent coloring is not available in my Stata SE 13.1.

Hi Nick, thanks for suggesting to use qgamma. Results below hint that it may not be a good fit? I want to check Weibull fit for the data as well but can't find out online how to get its beta/gamma parameters as used in your article 'Speaking Stata: Density probability plots' here https://www.stata-journal.com/sjpdf....iclenum=gr0012

[Nick Cox]

A Weibull distribution may be fitted using weibullfit from SSC and the corresponding quantile plot is at qweibull

The data you show would seem hopeless for any but a three-parameter gamma, which is more awkward to fit.

I feel Darwinian about density probability plots. They haven't survived in my data analysis practice.