PLS example

library(tidyverse)
library(mdatools)

Predictors train-set

dfXtr = read.csv('Colorigrammi_lambrusco_xl_TR_cleaned.csv', check.names = F, row.names = 1)
dfXtr = dfXtr[-1]
dfXtr

NA

Predictors test-set

dfXts = read.csv('Colorigrammi_lambrusco_xl_TS_cleaned.csv', check.names = F, row.names = 1)
dfXts

NA

Response train-set

dfYtr = read.csv('YAntocianiTR_cleaned.csv', check.names = F, row.names = 1)
dfYtr

NA

Response test-set

dfYts = read.csv('YAntocianiTS_cleaned.csv', check.names = F, row.names = 1)
dfYts

NA

PLS

m = pls(dfXtr[9:ncol(dfXtr)], dfYtr[ncol(dfYtr)], 
        cv = dfXtr$Replica.prelievo.in.campo,
        lim.type = 'jm', 
        center = T)

Take a look at the RMSE

par(mfrow=c(1,1))
plotRMSE(m)

NA
NA

We can choose 4 latent variables.

par(mfrow=c(2,2))

m = selectCompNum(m, 4)
plotXResiduals(m, norm = T)
plotPredictions(m)

plotPredictions(m$res$cv, cgroup = dfXtr$Tempi.di.prelievo, show.stat = T)
plotXScores(m$res$cal, cgroup = dfXtr$Tempi.di.prelievo)

par(mfrow=c(3,1))

plotXLoadings(m, type = 'l')
plotVIPScores(m, show.lines = c(NA,1), type = 'l')

plotRegcoeffs(m)

Now compare the test-set response values with the predicted ones (calculated on the x block of the test set )

res = predict(m, dfXts[9:ncol(dfXts)], dfYts[ncol(dfYts)])

plotPredictions(res, cgroup = dfXts$Tempi.di.prelievo, show.stat = T)