Dear All,
I would like to ask you about variation partitioning analysis by using vegan in R. Here is my data.
Partition of variation in RDA
Call: varpart(Y = myY, X = ~V10, ~V1 + V2 + V3 + V4 + V5 + V6 + V7 + V8 +
V9, data = myX, transfo = "hel")
Species transformation: hellinger
Explanatory tables:
X1: ~V10
X2: ~V1 + V2 + V3 + V4 + V5 + V6 + V7 + V8 + V9
No. of explanatory tables: 2
Total variation (SS): 1.3749
Variance: 0.12499
No. of observations: 12
Partition table:
Df R.squared Adj.R.squared Testable
[a+b] = X1 1 0.54823 0.50305 TRUE
[b+c] = X2 9 0.97764 0.87703 TRUE
[a+b+c] = X1+X2 10 0.98426 0.82686 TRUE
Individual fractions
[a] = X1|X2 1 -0.05017 TRUE
[b] 0 0.55322 FALSE
[c] = X2|X1 9 0.32381 TRUE
[d] = Residuals 0.17314 FALSE
And, I got venn diagram like in attached photo. I would like to know if I can use this figure. Because, the first R2 showed negative (-). And also, the residual value is so big.
best regards
lynn
Hi, I also found negative R2 adjusted values in variation partitioning in partial RDA. Legendre (2008) [doi: 10.1093/jpe/rtm001] argued that "Negative values of Ra2 are interpreted as zeros; they correspond to cases where the explanatory variables explain less variation than random normal variables would."
I hope it helps!
Felipe
First, I would be concerned about the fact that you have a negative R2. I'm not sure what you are doing exactly, but you might want to look at Chapt. 5 of the book below which has R code but uses different packages. Beware though because I am assuming you are interested in variance components. If I am wrong, please ignore this comment. best wishes. BTW if R2 was an adjusted R^2 that can be negative.
http://www.amazon.com/Analysis-Experiments-Chapman-Statistical-Science/dp/1439868131/ref=sr_1_4?ie=UTF8&qid=1458708310&sr=8-4&keywords=John+Lawson
Hi,
I agree David, I also suspect that the reason behind negative R^2 is that these values are adjusted R^2 values that CAN be negative.
Yes guys...these values are adjusted R square.
The alignment of the result a bit shifted when she paste the result here.
To answer her question, yes we will obtained a negative adjusted R square value due to the nature it is calculated. In general, the adjusted R square value depend on the number of observations and number of predictors in the model or analysis. You did get a negative value since you only have 12 observation but high number of predictors. you might want to consider or recheck your partition.
Hope this info is helpful.
Thanks all for your valuable reply. Yeah, Kamarul Mansor, I will try to reduce the predictors. Thanks again for your helpful answer.
Regards
lynn
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package