Agriculture Reference
In-Depth Information
0.254 and, in the Tukey comparison, it is 0.737. In both cases, there
is no difference (P>0.05), but the probability is much higher with
Tukey's test, which is known for being particularly conservative in
declaring differences significant and this is reflected in the higher P
value.
Several other multiple comparison tests are available in Stata. In
addition to the unadjusted and Tukey's mentioned above using the
pwcompare
command, there are the Bonferroni's, Šidák's, Scheffe's,
Student-Newman-Keuls' (SNK), Duncan's, and Dunnett's tests.
Dunnett's is a special case of multiple range testing where one treat-
ment is compared to all others. This might be used, for example, where
a particular standard animal ration is compared to several other ration
formulations. The treatment with the lowest identifying number is
considered the standard for comparisons.
Post hoc multiple range tests are available in Stata as we have
seen above. These tests are also included with CRD (completely
randomized design) within the
oneway
command. Open the file
virustestinoc.dta, which is the absorbance reading of virus-infected
watermelon germplasm. Antigen/antibody testing can be highly
specific for detecting virus diseases in plants. Such colorimetric tests
rely on the absorbance of light at a specific wavelength to deter-
mine if an infection is positive. After loading the dataset, enter the
command
oneway
absorb trt,
bonferroni tabulate
This results in the following output:
1-5: |
Different |
virus |
inoculated |
watermelon | Summary of ELISA Absorbance value
germplasm | Mean Std. Dev. Freq.
------------+------------------------------------
PI 025 | 0.520 0.036 11
PI 026 | 0.506 0.033 11
PI 261-1 | 0.275 0.048 11
PI 528 | 0.237 0.025 11
Egun | 0.147 0.054 11
------------+------------------------------------
Total | 0.337 0.156 55
Search WWH ::
Custom Search