Partial least squares regression (PLS)

MinutePlot^® and R on the same data. MinutePlot version 1.0.0; R 4.5.2.

Dataset

• oliveoil (pls package) — Sixteen olive oils (Greek, Italian, Spanish): five chemical measurements (acidity, peroxide value, K232, K270, DK) and six sensory-panel scores (R: pls::oliveoil). Both tables are hosted; the entry predicts the sensory attribute yellow from the five chemical variables.
• Data format for MinutePlot: oliveoil_minuteplot.csv — arranged exactly as the app’s data template for this test expects (paste it into the app).
• Data format for R: oliveoil_chemical.csv and oliveoil_sensory.csv — the long-format file the R script reads (keep it in the script’s folder).
• R built-in: the script loads oliveoil from the pls package with data(); the two R files are its two tables exported as CSV.

Setup

In MinutePlot:

• Workbench → PLS → paste oliveoil_minuteplot.csv → Preview data → tick first column is sample names (the sample column) → response column yellow (Y); Acidity, Peroxide, K232, K270, DK are the predictors (X) → Confirm → components 3, Scale variables ticked, cross-validation and permutation off → Run PLS regression.

In R:

• Install once: install.packages(c("pls")); then run oliveoil_pls.R top to bottom (source it, or paste it into the console) with the R data file in the same folder.
• R 4.5.2 and the package versions in the page header; each script prints the versions it runs under.

The analysis:

• PLS1 of yellow on Acidity, Peroxide, K232, K270 and DK, 3 components, predictors and response standardised (scale = TRUE on both sides); MinutePlot: NIPALS; R: pls::plsr(..., ncomp = 3, scale = TRUE) with the kernel algorithm -- identical for a single response.
• The two R tables: oliveoil_chemical.csv is X (the five predictors); oliveoil_sensory.csv supplies Y (its yellow column). MinutePlot takes one table, so oliveoil_minuteplot.csv holds the five predictors with yellow beside them -- the chemical file alone has no response and cannot reproduce this entry.
• Cross-validation and permutation are off for this comparison (they are resampling procedures, not deterministic fits). Component signs are arbitrary and aligned before comparison.

Side-by-side results

Variance explained per model size (%)

X loadings

Regression coefficients, 3 components (MinutePlot: per original unit of X; R’s pls::coef with scale = TRUE: per SD of X -- MinutePlot × SD(X) is the same quantity)

Verdict

R script

oliveoil_pls.R — each script loads its data, prints its versions and the R-side tables shown above, in page order; no plotting.