+Keith McGuinness hahaha, I know. Recently, I did a course about Numerical Ecology and I saw Primer very much, although I'm still needing many more courses to understand it better.
HI Keith, very nice video, but there is something that I can pick up very well. What I have learnt so far by watching all these video, is that with Permanova we get the same results with Minitab except the fact that, the pvalue on Permanova comes from the number of permutation, and that if we want to obtain the right p(value) as see on minitab, we have to click on Do Monte Carlo test on Permanova. But now, my question is what is the importance of increase the number of permutation? or to do the Pairwise test? I am lost there.Because at the end, the results change every time, and how to explain them, or how to present them?
Any procedure which involves permutation or randomization will probably give slightly different results each time it is run. Increasing the number of permutations (randomizations) increases the precision of estimates of the p-value (with one important qualification: see later). With 999 permutations, the smallest possible p-value is 0.001: most actual p-values smaller than this will (probably) show as 0.001. The important qualification is that this depends on the number of possible permutations: with few groups, and few replicates, the number of possible different permutations of the data may be smaller than 999. In this situation, PRIMER/PERMANOVA offers an alternative way to generate p-values: Monte Carlo. The alternate procedure should probably only be used when the number of different permutations is small: the manual has more information on these procedures, and their assumptions and limitations. The pairwise tests use subsets of the data so there may be enough different permutations for reliable p-values in the PERMANOVA but not in the pairwise tests. In such situations, the pairwise tests might have to be done using p-values generated by the Monte Carlo procedure.
You need to have the worksheet set up correctly. First row is variable labels; first column is row labels. One blank row after the last variable and then columns for any factors. One blank row after the last data row and then rows for indicators.
Thanks Keith for nice explanation, I am wondering about the estimates of variation, why it is not 100% (Residual and environmental factors combined).
If the design is unbalanced, variation may not add to 100% because of confounding. A complete explanation is too long to go into here.
Nice video, Keith.
+Renato Correia Thanks! That was quick!
+Keith McGuinness hahaha, I know. Recently, I did a course about Numerical Ecology and I saw Primer very much, although I'm still needing many more courses to understand it better.
HI Keith, very nice video, but there is something that I can pick up very well. What I have learnt so far by watching all these video, is that with Permanova we get the same results with Minitab except the fact that, the pvalue on Permanova comes from the number of permutation, and that if we want to obtain the right p(value) as see on minitab, we have to click on Do Monte Carlo test on Permanova. But now, my question is what is the importance of increase the number of permutation? or to do the Pairwise test? I am lost there.Because at the end, the results change every time, and how to explain them, or how to present them?
Any procedure which involves permutation or randomization will probably give slightly different results each time it is run. Increasing the number of permutations (randomizations) increases the precision of estimates of the p-value (with one important qualification: see later). With 999 permutations, the smallest possible p-value is 0.001: most actual p-values smaller than this will (probably) show as 0.001.
The important qualification is that this depends on the number of possible permutations: with few groups, and few replicates, the number of possible different permutations of the data may be smaller than 999. In this situation, PRIMER/PERMANOVA offers an alternative way to generate p-values: Monte Carlo. The alternate procedure should probably only be used when the number of different permutations is small: the manual has more information on these procedures, and their assumptions and limitations.
The pairwise tests use subsets of the data so there may be enough different permutations for reliable p-values in the PERMANOVA but not in the pairwise tests. In such situations, the pairwise tests might have to be done using p-values generated by the Monte Carlo procedure.
Is the manual available in pdf?
Yes, if you have the program.
Thank you. I am a beginner, I have a problem with importing data from excel
You need to have the worksheet set up correctly. First row is variable labels; first column is row labels. One blank row after the last variable and then columns for any factors. One blank row after the last data row and then rows for indicators.