Hi,
I have to compare means of one standard group with 5 reference groups. So my options are:
1) either i can compare reference groups one by one to the standard group by using T test
2) or i can compare means of reference groups together using ANOVA.
What is preferable and what would be the difference in results?
Hello Hasan,
Dunn's test is a special case of protected multiple-t tests (K - 1 groups vs. a comparison/standard/baseline group). It's available in SPSS and other anova software programs.
The final answer should be based on what your research question is.
Good luck with your work!
Dear Hasan,
In cases where you have more than two groups, you have to use any statistical method that possibility compares all the groups at ones (ANOVA is one example of these methods). This is recommendable because with these you control the type 1 error, and do not inflate it.
t-test and ANOVA- one factor and two modalities give identical results in hypothesis testing.
As David and Pablo wrote you must to perform ANOVA joint with multiple comparison test, first because you reduce the chance of type I error. If you use one by one T test you use only the variance estimation of the two groups in each test. Instead, ANOVA with multiple comparison test use the complete variance estimator of the total of your responses (for the 6 treatments) simultaneously in each paired comparison. The T test for ANOVA is called LSD (least significant difference) because the real p is greater than the supposed p. As example if you want p0.05 and it is possible that your test present type I error. Dunnet test for 1 treatment as control or Tukey test for complete multiple comparison are more proper than LSD test (or T test).
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA