I have two independent factors which are preservative concentration (0.5, 1.0, 1.5, 2.0 M) and days of treatment (3, 7, 21 days). The dependent factor is the mass loss after fungal decay. I also have a control that I did not expose to any treatment. What is the appropriate statistical test for my study?
My reviewer asked for statistical analysis in relation to control (Duncan's test). I don't understand what he means.
Duncan's test is a multiple comparison procedure. The result of the test is a set of subsets of means, where in each subset means have been found not to be significantly different from one another. It is especially protective against false negative (Type II) error at the expense of having a greater risk of making false positive (Type I) errors.
You can go with the test suggested by reviewer or you could go with ANOVA.
I recommend two way ANOVA of repeated data and after you use the new statistic of Cumming.
The significance p has limitation. SEE CUMMING statistic
https://thenewstatistics.com/itns/esci/
https://www.youtube.com/watch?v=iJ4kqk3V8jQ
http://www.psychologicalscience.org/members/new-statistics#p3
Depends on how you ran the experiment. I would take all my wood and divide it into four stacks. One stak would get no treatment. The remaining pieces would be evenly split between the 4 concentrations. After 3 days I would remove 1/3 of the wood. Half of the remaining pieces would be removed on day 7, and the rest are removed on day 21. All pieces would then be subjected to the rot treatment. Now there is a choice. I could assess each piece on day 10, 20, 30 ... end. This is a repeated measures design. However, the assessment might be destructive, so the piece measured on day 20 was not the one measured on day 10. This is not a repeated measures design. I could try to avoid the problem by only measuring rot on day 50. With the repeated measures design, I need to check for correlation between the results on day 10 versus day 20, day 20 versus 30, and so forth. If there is no correlation, then this is not a repeated measures design.
Duncan's test is one option. You could consider multiple regression. Wood loss = concentration + duration + error. The latter approach would let you guess as to the result of using 1.75M for 15.2 days.
You probably did not measure your control repeatedly (day 3, 7, 21). I would suggest you run 3 one way ANOVA's for each day, with the control as additional level of concentration. So, each time with 5 levels: No, 0.5, 1.0, 1.5, and 2.0 M preservative concentration. In these 3 ANOVA's, you each time could add the Duncan test, which should each time reveal which of the 4 concentrations are different from the control.
See attachment for the SAS program example.
I had assumed that the wood treatment was not entirely a surface treatment. Thus, the 3, 7, and 21 days were the length of time spent in the different preservative concentration solutions. The other issue is that I would not expect significant degradation of the wood in 3 days, and I would be surprised if there was measurable decay at 7 days. For fungal decay of wood, even in a tropical environment like Malaysia and depending on the type of wood, a few weeks to a few months for the first observation seems more reasonable?
Thank you for all feedback
I have understand the Duncan's test. It is one of the post-hoc test to run after your independent variables found to be significant and you want compare between means which of the pairs are significant and insignificant. I have corrected my paper, and it has been accepted for publication.
I suggest Tukey test... After an ANOVA test, try to group the treatments with Tukey test and for control you can use Dunnett test, which will compare your results with control. You can do it on SPSS software.
Greetings.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis