Some NAs produced during rda analysis in the contrained variables section

Question

I am running a RDA analysis (vegan package) following the next script:

env: The "env" dataframe has 9 variables: HIX, SR, FI, BIX, perC1, perC2, perC3, perC4, CDOM_S. This dataframe is standarized from 0 to 1 by each variable and does not have any NA values

#Running the RDA analysis:

rda1<-rda(bio ~ .,env)

scrs <- scores(rda(bio ~ ., env),scaling = "sites", correlation = TRUE)

scrs

coef(rda1)

If you noticed, the perC5 variable was removed! So, my question is, Why RDA analysis remove this variable?. I would like to keep all of my 9 variables in the analysis...

Thanks a lot in advance!

Answer 1

It was a five-class factor, and any factor with five classes can be expressed with four contrasts. Think about it: two classes (yes/no, TRUE/FALSE, control/treatment) is just one difference and for any factor variable, there is one less contrast than the number of classes. Don't worry, all levels were in your analysis, but they could be expressed with four contrasts. The behaviour is exactly the same in all standard R methods, such as in linear models (lm) or generalized linear models (glm).