I have performed a pot experiment for effect of AMF and PGPR on drought stress. I want to analyse the data by One way ANOVA. I have seen in most of the papers they use either Ducan test or Tukey's test in Post hoc. I wanted to know the difference between the two and which would be ideal for the study.
Both DMRT and tukey's tests are used to differentiate the significance of treatment means. I feel both are based of LSD (least significance difference) and are good to use. I suggest you use DMRT test as it widely being used in research and publications.
The method of Duncan can be applied in the same conditions of the Tukey's test, but with a significance (experiment wise) less cautionary and then a power to a single test (or comparison wise) greater. Consequently, makes significant differences between pairs of individual medium that are not those with the Tukey test, making for this preferred by various researchers, particularly in recent years; others continue to prefer Tukey, precisely because it is more conservative and therefore "less objectionable", once demonstrated the significance of a difference between two means.
The Tukey's post test is robust and provides correct results even if some of the ANOVA assumptions are violated. Tukey and Bonferroni's post test are strong and powerful test. Good statistical analysis avoids that experiments repeated from other groups obtain different results or are not confirmed. Wrong statistical analysis is not good for scientists and not for science.
Tukey is the best for all-possible pairwise comparisons when sample sizes are
unequal or confidence intervals are needed; very good even with equal
samples sizes without confidence intervals
Duncan use for comparing one sample (“control”) to each of the others, but not comparing the others to each other
Regards to the previous answers, I think that, in Duncan test there is no need for control treatment and there is no need for checking significant differences among the treatments (F value). Easily all the treatments could be compared, however, Tukey test is like LSD test and when we could this test that the F is meaningful and means are compared with a constant value.
Due to the lack of robustness of Duncan's test there are some journals that do not accept it as a proof of differences
Two articles. The older focuses on circumstances Duncan should not be used at all
http://fshs.org/proceedings-o/1976-vol-89/251-253%20(CHEW).pdf
the more recent, more generally on how when & how Duncan and uncorrected Fischer LSD have led to unacceptable / inaccurate conclusions throughout the literature.
Article Consequences of Common Data Analysis Inaccuracies in CNS Tra...
This does not specifically answer your questions RE which journals don't accept the test (presumably that's known by word of mouth or via a bad review), but it does outline why & when journals are likely to look askance at its use. It took about 3 minutes to google, BTW, using the search term, [duncan's test not accepted journals].
Thank you so much Joy Sabl, for your prompt and relevant explanation. I will go through the link you suggested.
I guess that Duncan`s test works well up to 25 mean values, Tuckey`s fitts more numbers.
- Badges
- Science topic
- Similar topics
- Filing