Hello,
I would like to reinforce my anova results using PCA i.e. which factor are most
important because they explain most of the variance (i.e. signal) of my 2^k*r
experiment. However, I get the following error while trying to run PCA:
> throughput.prcomp <-
prcomp(~No_databases+Partitioning+No_middlewares+Queue_size,data=throughput)
Error in prcomp.formula(~No_databases + Partitioning + No_middlewares + :
PCA applies only to numerical variables
What is the most R-like concise way to map/transform those factor values into
numerical values in a suitable way for PCA analysis? My first attempt would be:
# C++ "style"
throughput$No_databases_num <- (throughput$No_databases == 1) ? -1 : 1
throughput$Partitioning_num <- (throughput$Partitioning ==
"sharding") ? -1 : 1
etc.
How can I do this in the R way?
Would these -1, 1 be sensible for a PCA analysis or it just doesn't matter?
How about a factor for which I have 3 levels? -1, 0 and 1?
Many thanks in advance,
Best regards,
Giovanni