Agriculture Reference
In-Depth Information
are displayed next to the appropriate mean ( depend ) in descending
order. The program then restores the dataset back to its original state.
This program represents one method of solving this problem and in
programming there are often several different methods that can be used
to arrive at the same solution. In programming, it is often the marginal
case that gives the most trouble. For example, in the statement
local alph`f' = “`alph`f''” + substr( “abcdefghijklmnop
qrstuvwxyz” , `b' , 1 )
the number of letters available includes the entire alphabet. What if,
although unlikely, the number of differences exceeds this? This con-
dition could result in an error stopping the program. Programs that
are to be distributed as ado files and act like a built-in function often
require a great deal of programming to handle these marginal cases.
Programs that only you will use won't require this kind of rigor. It is
important and cannot be emphasized enough, however, the need to
document your code as you build it and use macro names that give
some idea of what it is. The Scheffé Test is considered a rather con-
servative test that many consider overprotecting against type II errors
(accepting the null hypothesis when the alternate hypothesis is true).
Duncan's New Multiple Range Test (MRT) was developed by the
statistician David Duncan in 1955 (I guess the New could be dropped).
Actually the New was added to distinguish this test from a previous
one proposed by Duncan. Duncan's MRT is a modification of the
Student-Newman-Kuel test that adjusts the alpha level based on the
distance of the treatment means from each other. Unlike the Scheffé
Test, which is often considered too conservative, Duncan's MRT
is often considered too liberal in declaring two means as different.
Unlike Scheffé Test that uses a single value to compare the treatment
means, Duncan's MRT uses different values to compare treatment
means based on how far apart the treatment means are when ranked
in descending order. So, for example, 10 treatments where the highest
treatment mean was compared to the fourth largest treatment mean
would use a different value for comparison than the highest value
compared to the sixth largest mean.
Load the dataset watertrial2007frtchar2.dta and open the do-file,
duncan.do. We will be using these files to examine Duncan's MRT.
Search WWH ::




Custom Search