Post Hoc Tests - Agricultural Statistical Data Analysis Using Stata - page 163

This do-file requires an ado file that is not part of the official package
of Stata ado files. The ado file in question is qsturng. To find this pro-
gram, enter in the Command window while connected to the Internet

findit qsturng

This command will open a Viewer window with a list of several Stata
Technical Bulletins (STB). Download dm64, which is in STB-46,
and install this ado command. qsturng stands for q studentized
range, which is found in tables at the back of many statistics textbooks.
This command requires three inputs: the number of treatments, error
degrees of freedom, and probability rate. For a 5% probability, enter
the value 0.95 and for a 1% probability enter 0.99.
The duncan do-file is almost identical to the scheffe do-file in terms
of presenting the results. It differs in the comparison values used to com-
pare means. With the Scheffé Test, a single test value is used, whereas
with Duncan's MRT there is a different value for each comparison.
Look at the segment of code below to see how this is calculated.

local var = (e(rmse)) ^ 2 // Error mean square from ANOVA

local repl = e(df_2) + 1 // Number of replications

local sd = sqrt( 2 * `var' / `repl' ) // Standard error of the mean
difference

local trt = e(df_1) + 1 // Number of treatments

forvalues x = 2 / `trt' { // Loop from 2 to number of
treatments

quietly : qsturng `x' e(df_r) 0.95 ^ (`x'-1)

/* Calculates the Studentized Range (this function is not part of
the official Stata ado files and will have to be downloaded */

local stu`x' = $S_1 //qsturng saves its results in
global S_1 & $S_1 returns the
value

local r`x' = `stu`x'' * `sd' /sqrt( 2 ) /* Using the Studentized Range
to calculate the significant
difference based on rank order
distance */
} // End 'x' brace

From the previous ANOVA, several e() scalars are available to use
in the program. The e(rmse) is the root mean square error, which when