I have a set of data from wounds that are going through normal wound healing and those that have been treated to intentionally delay wound healing (n=3) in pigs. For normal/acute wounds, tissues were harvest at two timepoints: Day 5 and Day 10 from the point of wounding. For wounds with delayed healing, four timepoints were collected: Day 5, 10, 15 and 20 from the point of wounding. These wounds are all on the same pig and repeated twice on two other pigs. I would like to answer the question: Is wound healing significantly delayed in the treated wounds compared to acute wounds.
Initially, I just used an ANOVA to compare across all wounds and all timepoints but I'm not sure if this is the appropriate test to use. Should I be doing a pairwise comparison for each timepoint (eg. Acute D5 V Treated D5) instead of analyzing all the timepoints together (as acute does not have timepoints day 15 and day 20).
Ηι Mandy Tan !
First of all, I would like some clarification about your variables
> V1: wound status (either treated or acute)
> V2: delayed would healing (logical, yes or no)
> V3: day when the wound was healed (the day when you removed the tissue and observed that the wound was healed)
I assume you somehow count how much time did it take for the would to heal? Do you have a variable for this?
In addition, I would need some clarification about the timestamps you remove the tissue.
You remove the tissue, you check whether the would is treated or not, right?
If I understand correctly, you should do ANOVA for the timestamp (V3) based on the the would status (V1). I assume that your data does not have a lot of records (if n=3 refers to pigs, they are not enough). Therefore, your results might be biased, or not statistical significant.
You can also try Kruskall Wallis Test, an alternative to ANOVA, that does not need any assumpions (i.e., normal distributed data).
You say you "have a set of data". What kind of data? What was measured, how was it measured? What is a reasonable distribution model or a random variable modelling this "data"?
Your time-points are very limited to estimate a time-course of wound-healing. Without much of an understanding it seems to me that this is a rather suboptimal expermental design to answer your question.
So far I's guess that it would be best to compare the data between "acute" and "delayed" obtained at day 5, and the same at day 10. If in both cases the data from "delayed" indicates a lower wound-healing, this may indicate that the healing is delayed. The problem might be that at day 5 the healing process is not far enough so that a difference can be recogniced, in which case the test at day 5 would be "not significant". I think this is tolerable as long as the test at day 10 is "significant". The other way around (significant at day 5 but not at day 10) would be "inconclusive evidence" regarding the delay.
If the acute healing process is essentially finished at day 10, you can use the days 15 and 20 from the delay-group to check if there is a healing process, althogh at day 10 there might not be much to see in the delay-group. You can also use a test of the monoton increase (testing Spearman's rho) in the delay group to show that there is a healing process (the same could be done in the acute-group, but here you can only compare day 10 to day 5).
You need to do some type of repeated measures analysis to account for the changes in individual wounds over time. This will give you more statistical power. Here is an article specific to methodology for wound healing, which is probably non-linear (assuming you're even measuring it numerically, such as by area). http://www.newjerseywoundhealing.org/Resources/statistical-analysis-of-wound-healing-rates-for-pressure-ulcers.pdf
With an "n" of three, I don't think that you can do any statistical test that will provide a probability on a population-based assumption. The best that you can do is to graphically demonstrate the trends in the data over the duration of your data collection but as for predicting the impact of your procedure - there is insufficient data to draw any robust statistically valid conclusions.
More samples + Test for normality and homogeneity, then consider ANOVA. Now, it is better to use the case study method.
if, I understood correctly, you want to compare the healing time between two different groups, in this situation you have two means, and if you had to test the normality of the distribution, the student's T test is appropriate, on the other hand if the distribution does not follow the normal law, it is the mann - whitney test , on the other hand if you have more means and the distribution follows the normal distribution, it is the ANOVA test, and if the disribution of the variable does not follow the normal distribution, it is the H- test of Kruskal wallis
All statistical tests are appropriate. What varies is in which scenario.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics