While comparing several biogeochemical parameters measured in different stations of the same estuary, or two different aquaculture ponds within the same wetland, reviewers have asked for carrying out repeated measures ANOVA instead of students t-test. I would like to know more about the applicability of repeated measures ANOVA, like when can it be applied? why is it better than students t-test (independent samples)? and most essentially when can we not consider this test?
Hello Abhra,
The distinction between opting for an independent batches test vs. a dependent / correlated / matched batches test hinges on how data are collected and recorded. If you were to compare temperature measures from persons, one taken orally vs. one taken axially (e.g., underarm), these are very likely to exhibit a correlation, since persons with higher temperature are likely to show higher values both ways, and persons with lower temperature are likely to show lower values both ways. Hence, a dependent t-test (or ordinary repeated measures ANOVA) would be preferred as the analytic method.
What happens if you were to use an independent t-test (or ordinary one-way, between-subjects ANOVA) instead would be that your error term would be larger, resulting in a less sensitive test of whether site/method for temperature matters. As long as the matched/paired case scores are positively correlated, the standard error of difference scores (the "error term") will be smaller than that for the independent groups approach.
So, the question for your research is, are the scores from different stations in the same estuary likely to be correlated? If so, dependent-t or repeated measures anova is one way to go. However, a competing approach (especially if you had estuaries within wetlands, or something comparable) would be a multi-level model (sometimes referred to as a hierarchical linear model) instead. Without knowing more about what data points you have, and how data were collected, it's difficult to offer an emphatic choice.
You may wish to consult similar published studies to see how other researchers have opted to treat their data. Of course, whatever choice you make, you'll need to justify that for your own reviewers and target audience.
Good luck with your work.
Now that was such a lucid explanation and so easy to understand. You cleared many of my doubts, Sir. I really appreciate it.
Here is a useful link to guide you for more hint on the repeated measures anova.
https://statistics.laerd.com/statistical-guides/repeated-measures-anova-statistical-guide.php#:~:text=Repeated%20measures%20ANOVA%20is%20the,or%20ANOVA%20for%20correlated%20samples.
Best of luck.