I am analysing data obtained from experiments conducted on a group of 10 individuals in triplicate. In each replicate, the number of data points are not necessarily the same for each of the individuals and such variation can also be observed between replicates. This is due to the nature of the procedure, i.e. RAPD PCR.
You can use random effects ANOVA for possibly unbalanced data, however you need an idea about vthe structure of your data. (which is not clear for me from your description). What are fixed effects, what are random effects? From variance components, you can derive reproducibility, or between-subject, and within-subject variability. I recomment to use a table with a column for each factor (group, subject, replicate etc.), and one column for the values.
In some statistical textbooks formulas are available. Today, statistical software (SAS, R, ...) is the best way to perform these analysis, however, one has to understand the structure of the data.
I would be wary of an ANOVA. I would use an ANCOVA. this appears to be stochastic data.
I have conducted amplifications from 10 samples of template DNA using 20 RAPD primers and I have repeated this 3 times for each primer. The data for each sample consists of a number of PCR bands of varying size (i.e. a genetic profile). Within each replicate, some bands will be shared by all 10 samples, some bands will be shared by some samples but not others and some bands will be unique to certain samples. The same will apply to each sample across replicates. I can see 'by-eye' that some primers are more reproducible than others but I would like to draw my conclusions in a more objective manner. The data will ultimately be binary in nature.
If I was dealing with one replicate, the approach would be very straight-forward to me: the number of bands replicated as a percentage of the total number of bands produced. The three-replicate scenario is what confuses me. For 1 sample, if I have 5 bands in my first experiment, 3 of those are repeated in my second and 4 are repeated in my third. what is the overall reproducibility? And is there anyway that I can average this across all 10 samples so that I have one measure of reproducibility for each primer?
After you get the gel images in RAPD experiments, selecting correct analytic algorithms become the key for the data analysis. Most likely RAPD data will not be binary and desity-unweighted algorithms may not be suitable to use. Please see my paper in ResearchGate. The desity-weighted algorithms that we described in the paper are capable of handling comparisons of more than 2 samples.
Suitability and Advantages of Abundance-Weighted Algorithms for the Holistic Comparison of Sample Systems. J Appl Biotechnol Bioengineering, 2017, 2(6): 00050
- Badges
- Science topic