[R] LDA variables dropping similar within group

[Marina Fernandez]

Dear R-users:

I am R-usser begginer and at the same time a beginner running 
discriminant analysis;
I wanted to perform a DA using just the 80ß% of the original data but I 
have some problems with simmilarity in variables,

Here my Skript....


set.seed(123)
data80 <- data[sample(472, 378), ]
data80

#Remove all missing values listwise
data80.withoutna<-na.omit(data80)

#Group variable
data_grouping80<-data80.withoutna[,2]
data_grouping80

#dim(data80)

#Possible independent variables
variables80<-data80.withoutna[,4:447]
variables80


#Data set for discriminant analysis
ldadataset80<-cbind(data_grouping80,variables80)
ldadataset80

#Discriminant analysis as SPSS does it (excluded variables by SPSS, 
denoted by -)
library(MASS)

model_lda80<-lda(data_grouping80 ~. ,data=ldadataset80, 
prior=c(255/471,100/471,76/471,40/471))


model_lda80<-lda(data_grouping80 ~. 
-CHLOSTA-DIGGRAN-DRYFILMA-EQUSYLV-EUPALPI-GERPHAE-GERROBE-HYPORAD-JUNCOMNA-JUNEFFU-JUNFILI-JUNJAQU-JUNTRIF-KNAARVE-KNAMAXI-KOBMYOS-KOEHIRS-KOEPYRA-LASHALL-LASKRAP-LATLAEV-LATPRAT-LEOHISP-LEUALPI-LEUVULG-
LILMART-LINCART-LISOVAT-LOIPROC-LOLPERE-LOTCORSL-LUZCAMP-LUZLUTE-LUZLUZO-LUZPILO-LUZSPIC-LUZSYLV-LYCALPI-MAIBIFO-MELPRAT-MELSYLV-MENAQUA-MINGERA-MOECILI-MOLCAER-MUTADON-MYOALPE-MYOARVE-MYODECU-MYOSCOR-
NARSTRI-NIGRHEL-ONOMONT-OREDIST-OXYCAMP-PARLILI-PARPALU-PEDELON-PEDFOLI-PEDROSC-PEDTUBE-PEDVERT-PELAPHT-PERBIST-PERVIVI-PEUOSTR-PHLCOMM-PHLPRAT-PHLRHAE-PHYBETO-PHYHEMI-PHYORBI-PHYOVAT-PICABIE-PIMMAJO-
PIMSAXI-PINCEMB-PINMUG-PINVULG-PLAALPI-PLAATRA-PLABIFOL-PLALANC-PLAMEDI-PLESCHR-POAALPI-POAAMAR-POAANN-POAPRAT-POASUPI-POATRIV-POAVARI-POLALPE-POLAMAR-POLCOMO-POLJUNI-POLVULG-POTANSE-POTAURE-POTCRAN-POTEREC-
POTGRAND-PRIAURI-PRIELAT-PRIFARI-PRIMINI-PRIVERI-PRUGRAN-PRUVULG-PSEALBI-PULALPAL-PULALPAP-PULANGU-PULVERN-PYRCHLO-PYRMEDI-RANACON-RANACRI-RANBULB-RANMONT-RANNEMO-RHIALEC-RHIGLAC-RHIMINO-RHOFERR-RHYSQUA-
RHYTRIQ-ROSPEND-RUMACELL-RUMACET-RUMALPE-RUMALPI-RUMOBTU-RUMSCUT-SAGINASP-SALAURI-SALHERB-SALRETI-SALRETU-SALVPRA-SANMINO-SANOFFI-SCACANE-SCACOLU-SCALUCI-SCOAUTU-SCOHELV-SCOHUMI-SCOMONT-SELSELA-SEMARAC-
SEMMONT-SEMWULF-SENABRO-SENDORO-SENINCA--SESALBI-SIBPROCU-SILACAU-SILDIOI-SILLATI-SILNUTA-SILVULG-SOLALPI-SOLMINI-SOLPUSI-SOLVIRG-SORAUCU-STEGRAM-STEMEDI-TARALPI-TAROFFI-THAAQUI-THEALPI-THYPRAE-THEPYR-
THYPULE-THYSERP-TOFCALY-TRAGLOB-TRAPRAT-TRIALPE-TRIALPI-TRIBADI-TRICESP-TRIFLAV-TRIMEDI-TRIMONT-TRIPRAT-TRIREPE-TROEURO-URTDIOI-VACGAUL-VACMYRT-VACVITI-VALMONT-VALOFFI-VERALBU-VERALPI-VERBELL-VERCHAM-
VERFRUT-VEROFFI-VERSERP-VICCRAC-VIOBIFL-VIOCANI-VIOHIRT-VIOTHOM-VIOTRIC-WILSTIP
,data=ldadataset80)

##New variables# (variables 82 103 128 146 181 appear to be constant 
within groups)

#####I got as an answer that some variables are constant within groups, 
so I delete them fro the data as follows

set.seed(123)
data80 <- data[sample(472, 378), ]
data80
newdata80 <- data80[c(-82,-103,-128,-146,-181)]
newdata80

#####Then I computed the whole analisis again, but then i got the same 
answer at the end, just in this case the variables are different..


#Remove all missing values listwise
newdata80.withoutna<-na.omit(newdata80)
newdata80.withoutna
#Group variable
ndata_grouping80<-newdata80.withoutna[,2]
ndata_grouping80
dim(newdata80)
#Possible independent variables
nvariables80<-newdata80.withoutna[,4:442]
nvariables80

ldadatasetn80<-cbind(ndata_grouping80,nvariables80)
ldadatasetn80

library(MASS)

model_ldan80<-lda(ndata_grouping80 ~. 
-CHLOSTA-DIGGRAN-DRYFILMA-EQUSYLV-EUPALPI-GERPHAE-GERROBE-HYPORAD-JUNCOMNA-JUNEFFU-JUNFILI-JUNJAQU-JUNTRIF-KNAARVE-KNAMAXI-KOBMYOS-KOEHIRS-KOEPYRA-LASHALL-LASKRAP-LATLAEV-LATPRAT-LEOHISP-LEUALPI-LEUVULG-
LILMART-LINCART-LISOVAT-LOIPROC-LOL.......,data=ldadatasetn80)


Were an I falling? I can´t understand this seceond answer wit new 
similar variables when I alreday drop the ´variables that initially were 
similar within groups ones said ..

Thank you very much in advance!!!!

Kind regards