Dunnett's Test Demo

The following six waves represent results of five different treatments (data2-data6) and a control (data1) to six objects (rows).

data1data2data3data4data5data6
39.9728.1240.8246.5155.7945.71
40.4733.8343.8541.853.731.78
37.6835.6246.544.0749.5149.01
37.0434.9240.0848.1162.342.37
43.5828.7139.0943.0450.3943
41.8230.9437.4539.9856.3940.39

The Dunnett test compares the mean of each wave with that of the control. H0 depends on the /TAIL flag which by default corresponds to testing equality of means. To run the test execute the command:

StatsDunnettTest/T=1/CIDX=0/TAIL=1  data1,data2,data3,data4,data5,data6

The results of the operation are displayed in Dunnett Test table:

ComparisonDifferenceSEqqpConclusion
data1 vs data28.073.203652.5192.335180
data1 vs data3-1.2053.20365-0.3761332.335181
data1 vs data6-1.953.20365-0.608682.335181
data1 vs data4-3.8253.20365-1.193952.335181
data1 vs data5-14.58673.20365-4.553142.335181

In this case, we accept H0, i.e., the mean value of the control wave (data1) is less than or equal to the mean values of waves: data3, data6, data4 and data5. To test H0 corresponding to a control mean that is greater than or equal to other wave means execute the command:

StatsDunnettTest/T=1/CIDX=0/TAIL=2  data1,data2,data3,data4,data5,data6

The results are given again in the Dunnett Test table:

ComparisonDifferenceSEqqpConclusion
data1 vs data28.073.203652.5192.335181
data1 vs data3-1.2053.20365-0.3761332.335181
data1 vs data6-1.953.20365-0.608682.335181
data1 vs data4-3.8253.20365-1.193952.335181
data1 vs data5-14.58673.20365-4.553142.335180

Inspecting the "Difference" column, we can see that the waves data3, data6 and data4 have mean values that are close to the mean value of the control wave the other two samples data2 and data5 are included or excluded in the group depending on the tail being tested. To run the two tail test execute the command:

StatsDunnettTest/T=1/CIDX=0/TAIL=4  data1,data2,data3,data4,data5,data6

The results are given again in the Dunnett Test table:

ComparisonDifferenceSEqqpConclusion
data1 vs data28.073.203652.5192.656851
data1 vs data3-1.2053.20365-0.3761332.656851
data1 vs data6-1.953.20365-0.608682.656851
data1 vs data4-3.8253.20365-1.193952.656851
data1 vs data5-14.58673.20365-4.553142.656850

In this case we have a larger critical value which happens to accept the difference between the means of data1 and data2 but not data1 and data5.