Hi Keith, For my dissertation I need to compare invertebrate assemblages upstream and downstream of a river (Upstream reach is natural, downstream reach is channelized). Within each reach I've got 1 riffle and 1 glide. Which ANOSIM should I be running? One way or Two way? All results are significant, but the one ways (reach on their own, and meso habitat on their own) give R values around 0.2 and 0.35. On the two way the same factors give me 0.5 and 0.4
HI, Thanks for the video showing the difference between ANOSIM and Permanova. Very useful and interesting. I have a question, when I ran the one way ANOSIM for my zone(sites) and Location (Month), it showed that the was a significant difference among zone (Sample statistic (Global R): 0.046 Significance level of sample statistic: 0.6% Number of permutations: 999 (Random sample from a large number) Number of permuted statistics greater than or equal to Global R: 5) And that there was not a difference among location (Sample statistic (Global R): 0.009 Significance level of sample statistic: 19.6% Number of permutations: 999 (Random sample from a large number) Number of permuted statistics greater than or equal to Global R: 195) But when I did the two way ANOSIM with replication I obtained a different results showing that there is a significant difference there :TESTS FOR DIFFERENCES BETWEEN Month GROUPS (across all Site groups) Global Test Sample statistic (Global R): 0.205 Significance level of sample statistic: 0.1% Number of permutations: 999 (Random sample from a large number) Number of permuted statistics greater than or equal to Global R: 0 ( TESTS FOR DIFFERENCES BETWEEN Site GROUPS (across all Month groups) Global Test Sample statistic (Global R): 0.233 Significance level of sample statistic: 0.1% Number of permutations: 999 (Random sample from a large number) Number of permuted statistics greater than or equal to Global R: 0 The Permanova that I did, showed that there is significant difference among sites, not among month, and a difference between site x month : PERMANOVA table of results Unique Source df SS MS Pseudo-F P(perm) perms Si 9 941.71 104.63 5.6338 0.001 999 Mo 5 160.28 32.056 1.726 0.148 998 SixMo 45 3049.8 67.772 3.6491 0.001 997 Res 120 2228.7 18.573 Total 179 6380.5 If I am not wrong, I think the results from the Permanova is the best and I should keep that one, or maybe try to run the ANOVA and compare. Thanks once more
A two-way (or two-factor) design run through a one-way analysis can give misleading results, so is best not done unless there are special circumstances. ANOSIM in PRIMER 6 cannot test the two-factor interaction and this is a major limitation. ANOSIM in PRIMER 7 is better as it is more powerful. The PERMANOVA has a highly significant Si*Mo interaction, so an analysis excluding the interaction may be misleading. ANOSIM and PERMANOVA use different data. ANOSIM uses the ranks of the distances but PERMANOVA uses the actual distances. In some cases, this may cause different results. I would use the PERMANOVA unless there is a good reason to use the ranks (ANOSIM).
Thanks you so much Keith, You cannot know how much helpful are all your comments to me. I am really grateful for that. In another video, I asked you the question about the "No replication at the lowest level", After watching your video again and again, I understood and I noticed the mistakes that I was doing using my data. Thanks a lot. Unfortunately, I don't have Primer 7, and nobody around me has this version, so I can just use the version 6 that I have. Please Keith, concerning the Anosim test, is it just applicable to biological data? or we can also use it for environmental data? My other question is to know if there is a way on Primer to look at the relationship between different data that I have especially three types of data ( I have environmental data, biological data including Biomass and abundance). I have already look at the relationship between environmental data and biomass, between environmental data and abundance, so now I would like to know which environmental variable explain the change in the Biomass and abundance. Thanks once more
ANOSIM can be used for the environmental data and, as far as I know, can be used any time ordinary ordination (e.g. PCA, nMDS,PCO) is appropriate. For your last question, if you really mean 'which environmental variable(s)' then you would be looking at BEST in PRIMER and/or distLM in PERMANOVA+. These, however, take one set of predictor variables and one set of variables to be explained, so you would look at biomass and abundance separately.
Hi Keith, Thank you for your helpful video. In my analysis of comparing two different sites in terms of species composition, I have a negative R value returned to me, would you be able to explain what this means as I can't find any helpful info anywhere! Thanks again.
Taya Parker-Mian Apologies, for some reason I did not get a notification about your comment. R can fall below 0 by chance. From "Change in marine communities": "R substantially less than zero is unlikely since it would correspond to similarities across sites being higher than within sites; such an occurence is more likely to indicate an incorrect labelling of samples". The authors also note that Chapman and Underwood (1999) point out some situations (species poor communities with individuals heavily clustered) where R can be negative. C&U is in Mar. Ecol. Prog. Ser. 180:257-265.
Hello Dr Keith, Thank you for your video. For your information, I'm using the trial version of PRIMER 7 and I'm still very new to this software and ANOSIM. I want to perform ANOSIM using PRIMER 7 but I find that the window option are slightly different to PRIMER 6 as showed in your video. Anyway, I've manage to conduct ANOSIM for my data set which give me the result of R= 0.86 and significance level = 0.1% with number of permutation = 999. Do you think it is necessary for me to further reduce the significance level? and how could I do that. (there is no option for me to increase the number of permutation as you have shown). I'm looking forward to your answer and thank you very much!
There were interface changes between PRIMER 6 and 7. For ANOSIM, the permutations are set lower left in the dialogue box BUT the number of permutations may be limited by the data. A significance level of 0.1% equals a p-value of 0.001, which means that the R for the data (0.86) was bigger than all of the 999 permutations/randomisations. With 999 permutations, the smallest possible p-value is 0.001, or 0.1%. Increasing permutations will give a more accurate estimate of p.
Hello, I just wanted to state my appreciation for these videos. They are so clear and informative for a newbie like me. Thanks a million!
Hi Keith, this is great you are the only person on youtube doing PRIMER tutorials. Hero!
Thank you! I am happy my videos helped.
A massive thanks to you for your informative videos on PRIMER. Really has helped me a lot.
This is an incredibly helpful tutorial. Much better than combing through countless online fora. Thanks!
Thank you for a great tutorial!
Hi Keith,
For my dissertation I need to compare invertebrate assemblages upstream and downstream of a river (Upstream reach is natural, downstream reach is channelized). Within each reach I've got 1 riffle and 1 glide.
Which ANOSIM should I be running? One way or Two way?
All results are significant, but the one ways (reach on their own, and meso habitat on their own) give R values around 0.2 and 0.35.
On the two way the same factors give me 0.5 and 0.4
HI, Thanks for the video showing the difference between ANOSIM and Permanova. Very useful and interesting. I have a question, when I ran the one way ANOSIM for my zone(sites) and Location (Month), it showed that the was a significant difference among zone (Sample statistic (Global R): 0.046
Significance level of sample statistic: 0.6%
Number of permutations: 999 (Random sample from a large number)
Number of permuted statistics greater than or equal to Global R: 5)
And that there was not a difference among location (Sample statistic (Global R): 0.009
Significance level of sample statistic: 19.6%
Number of permutations: 999 (Random sample from a large number)
Number of permuted statistics greater than or equal to Global R: 195)
But when I did the two way ANOSIM with replication I obtained a different results showing that there is a significant difference there :TESTS FOR DIFFERENCES BETWEEN Month GROUPS
(across all Site groups)
Global Test
Sample statistic (Global R): 0.205
Significance level of sample statistic: 0.1%
Number of permutations: 999 (Random sample from a large number)
Number of permuted statistics greater than or equal to Global R: 0
( TESTS FOR DIFFERENCES BETWEEN Site GROUPS
(across all Month groups)
Global Test
Sample statistic (Global R): 0.233
Significance level of sample statistic: 0.1%
Number of permutations: 999 (Random sample from a large number)
Number of permuted statistics greater than or equal to Global R: 0
The Permanova that I did, showed that there is significant difference among sites, not among month, and a difference between site x month : PERMANOVA table of results
Unique
Source df SS MS Pseudo-F P(perm) perms
Si 9 941.71 104.63 5.6338 0.001 999
Mo 5 160.28 32.056 1.726 0.148 998
SixMo 45 3049.8 67.772 3.6491 0.001 997
Res 120 2228.7 18.573
Total 179 6380.5
If I am not wrong, I think the results from the Permanova is the best and I should keep that one, or maybe try to run the ANOVA and compare.
Thanks once more
A two-way (or two-factor) design run through a one-way analysis can give misleading results, so is best not done unless there are special circumstances.
ANOSIM in PRIMER 6 cannot test the two-factor interaction and this is a major limitation. ANOSIM in PRIMER 7 is better as it is more powerful.
The PERMANOVA has a highly significant Si*Mo interaction, so an analysis excluding the interaction may be misleading.
ANOSIM and PERMANOVA use different data. ANOSIM uses the ranks of the distances but PERMANOVA uses the actual distances. In some cases, this may cause different results.
I would use the PERMANOVA unless there is a good reason to use the ranks (ANOSIM).
Thanks you so much Keith, You cannot know how much helpful are all your comments to me. I am really grateful for that. In another video, I asked you the question about the "No replication at the lowest level", After watching your video again and again, I understood and I noticed the mistakes that I was doing using my data. Thanks a lot. Unfortunately, I don't have Primer 7, and nobody around me has this version, so I can just use the version 6 that I have.
Please Keith, concerning the Anosim test, is it just applicable to biological data? or we can also use it for environmental data?
My other question is to know if there is a way on Primer to look at the relationship between different data that I have especially three types of data ( I have environmental data, biological data including Biomass and abundance). I have already look at the relationship between environmental data and biomass, between environmental data and abundance, so now I would like to know which environmental variable explain the change in the Biomass and abundance.
Thanks once more
ANOSIM can be used for the environmental data and, as far as I know, can be used any time ordinary ordination (e.g. PCA, nMDS,PCO) is appropriate. For your last question, if you really mean 'which environmental variable(s)' then you would be looking at BEST in PRIMER and/or distLM in PERMANOVA+. These, however, take one set of predictor variables and one set of variables to be explained, so you would look at biomass and abundance separately.
Hi Keith, great videos! Question...how would you set up a 4 x 4 Latin Square in Permanova?
Hi Keith, Thank you for your helpful video. In my analysis of comparing two different sites in terms of species composition, I have a negative R value returned to me, would you be able to explain what this means as I can't find any helpful info anywhere! Thanks again.
Taya Parker-Mian Apologies, for some reason I did not get a notification about your comment. R can fall below 0 by chance. From "Change in marine communities": "R substantially less than zero is unlikely since it would correspond to similarities across sites being higher than within sites; such an occurence is more likely to indicate an incorrect labelling of samples". The authors also note that Chapman and Underwood (1999) point out some situations (species poor communities with individuals heavily clustered) where R can be negative. C&U is in Mar. Ecol. Prog. Ser. 180:257-265.
Thanks! Glad they are helpful.
Hello Dr Keith,
Thank you for your video.
For your information, I'm using the trial version of PRIMER 7 and I'm still very new to this software and ANOSIM. I want to perform ANOSIM using PRIMER 7 but I find that the window option are slightly different to PRIMER 6 as showed in your video.
Anyway, I've manage to conduct ANOSIM for my data set which give me the result of R= 0.86 and significance level = 0.1% with number of permutation = 999.
Do you think it is necessary for me to further reduce the significance level? and how could I do that. (there is no option for me to increase the number of permutation as you have shown).
I'm looking forward to your answer and thank you very much!
There were interface changes between PRIMER 6 and 7. For ANOSIM, the permutations are set lower left in the dialogue box BUT the number of permutations may be limited by the data. A significance level of 0.1% equals a p-value of 0.001, which means that the R for the data (0.86) was bigger than all of the 999 permutations/randomisations. With 999 permutations, the smallest possible p-value is 0.001, or 0.1%. Increasing permutations will give a more accurate estimate of p.
If you can't change permutations, that might be a limitation of the trial version. PAST also does ANOSIM.
PERMANOVA is a PRIMER add in made by the people who make PRIMER. So you need to buy it from them.
Hi Keith,
Where I can find the P value of ANOSIM result in Primer? Thanks
in your ANOSIM tab divide the value of "Significance level of sample statistic:" by 100
0.01 / 100 = 0.0001
the pvalue is = 0.0001
How can I add PERMANIOVA to my PRIMER? Thanks
You have to buy the add-in from the same company.