I got cells from different donors (n=5 minimum; n=12 maximum) and then those cells were cultured under four different conditions (A, B, C, D). I want to compare if there is any difference between the four conditions for a given parameter.
First question: Should I consider that the data is paired or unpaired? I am assuming that is paired..
Then I performed a normality test and the data is not normal. Thus, I performed a Friedman test followed by Dunn's multiple comparisons test.
As the n is small, and non-parametric test have low power, should I transform (log) the data to follow Gaussian distribution instead and then perform RM-one way ANOVA test followed by Tukey's multiple comparisons test?
How would you perfom the analysis?
Thank you in advance,
Daniela Silva
Yes, it's a repeated measures ("paired") design.
If it is reasonable to assume a log-normal distribution and that relative effects are morerelevant than additive effects, iyou should log-transform the data and analyze this with a two-way or repeated-measures ANOVA. The pair-wise comparisons wouldactually not be done with Tukey's procedure (to myknowledge this is not for paired data), but there are multiple testing corrections available.
Daniela Silva ,I don't think it is a repeated measure, but random samples of between s(5) to s(12)[unequal and small samples] distributed to the four independent conditions- clearly the data is NOT paired, except if you have a pre-value and post-value for analysis! For a given parameter, (say the mean [or median(non parametric))] of the response measure, what you need to test for is the normality, same sample sizes, significant outliers, and heterogenuity before using the One Way ANOVA, {which only require approximately normal data because it is quite "robust" to violations of normality, i.e, assumption can be a little violated and still provide valid results!} You can test for normality using the Shapiro-Wilk test of normality, and since the samples are small, a Q-Q or rather Normal Probability Plot is more APPROPRIATE!). *The Kruskal-Wallis H Test is the non-parametric alternative to a parametric ONE-WAY between groups analysis of variance (ANOVA) for independent groups when the sample size is small, normality assumption is failed, an outlier but not significant. Follow by the right pairwise post-test (on significance) pls. Alternatively, if your dependent variable [DV] is the time until an event happens(under the four cell conditions or treatments), then you might need to run a Kaplan-Meier Analysis. Hope these give some light, helpful. Silva ,
- Badges
- Science topic