# [R] Dunnett's T3 test (DTK package) problem

Sam Mak smak at gfz-potsdam.de
Mon Jul 1 14:45:01 CEST 2013

```Hello,

I would like to ask a question about using DTK.test() to do the T3 test.

http://ubuntuone.com/0zBl8QgUg1lAG6iK1CUicD

My question is: in the result of T3 test, is it possible for the following
three statements to hold without conflict?
1. m_a and m_b to be significantly different
2. m_a and m_c are not significantly different.
3. (m_c - m_a) >= (m_c- m_b)
where m_i is the mean value of group i.

I know that for Tukey test, the above three statements can't hold without
conflict.

The code to run the test is:

======== BEGIN ============
library(DTK)
DTK_result <- DTK.test(df\$measurement, df\$model, .05)
print(DTK_result)
======== END ============

Part of the result is displayed below. It says model05 and model01 are
significantly different, while model04 and model01 are not.

======== BEGIN ============
Diff    Lower CI  Upper CI
model02-model01 0.068468884 -0.04765040 0.1845882
model03-model01 0.143238920 -0.05589008 0.3423679
model04-model01 0.149002642  0.02900301 0.2690023
model05-model01 0.151504289 -0.04775209 0.3507607
model06-model01 0.193093625  0.07023934 0.3159479
...
======== END ============

The mean value of the groups are shown below. It shows (m_5 - m_1) > (m_4 -
m_1), where m_i is the mean value of model i.

======== BEGIN ============
library(plyr)
ddply(df, ~model, summarise, mean=mean(measurement))

model      mean
1  model01 0.5352273
2  model02 0.6036962
3  model03 0.6784662
4  model04 0.6842300
5  model05 0.6867316
6  model06 0.7283210
7  model07 0.7489938
8  model08 0.7941213
9  model09 0.7944798
10 model10 1.1399370
11 model11 1.2582504
12 model12 1.2928783
13 model13 1.5789970
======== END ============

--
View this message in context: http://r.789695.n4.nabble.com/Dunnett-s-T3-test-DTK-package-problem-tp4670649.html
Sent from the R help mailing list archive at Nabble.com.

```