Agriculture Reference
In-Depth Information
can be used; however, in this case, to conform to Zar's example, we
are using base 10 log. Now enter the command
tabstat measure tranmeas, statistics(mean sd var cv)
by( grp )nototal
This command calculates the mean, standard deviation, variance, and
CV for both the original data ( measure ) and the transformed data
( tranmeas ) as shown below:
Summary statistics: mean, sd, variance, cv
by categories of: grp
grp | measure tranmeas
---------+--------------------
1 | 3.28 .6306574
| .2863564 .0293007
| .082 .0008585
| .0873038 .0464605
---------+--------------------
2 | 6.94 .8988198
| .6024947 .0329628
| .3629999 .0010865
| .0868148 .0366734
------------------------------
The variances for the original data are obviously different (grp 1 = 0.082
versus grp 2 = 0.3629999) for each treatment and the standard devia-
tions are proportional to the means (grp 1 = 0.2863564 versus grp 2 =
0.6024947) resulting in coefficients of variation that are similar. After
transformation, the variances are homogeneous (grp 1 = 0.0008585
versus grp 2 = 0.0010865) and the standard deviations are not propor-
tional to the means (grp 1 = 0.0293007 versus grp 2 = 0.0329628).
Finally a log transformation may be used where the effect is multi-
plicative rather than additive. For example, in a RCBD (randomized
complete block design), it is assumed there is an additive treatment
and block effect. That is, from one block to another, the effect does not
change in orders of magnitude.
This is another case where we will go outside of Stata to find a
command. Enter findit nonadd in the Command window while
connected to the Internet. This will locate this command, nonadd ,
which can be downloaded and installed in Stata. Load the dataset
Onion Disease Transform Data.dta and enter the command
Search WWH ::




Custom Search