SPSS command: SPSS PLS

Partial least squares (PLS) regression in your browser

Partial least squares regression predicts one or more outcomes from many correlated predictors by first extracting a few latent factors that are built to explain the outcomes.

Run Partial Least Squares Regression in your browser All SPSS procedures

Free · No account · Files stay on your device · Updated 18 October 2026

What it tells you

Crosstabs runs SPSS-style PLS on your device with scale or categorical dependents, factors and covariates, and reports the proportion of variance explained per factor, variable importance in the projection (VIP), weights, loadings and coefficients in original units.

How much of the variance in the dependents (cumulative R² and adjusted R²) and in the predictors each additional latent factor explains, so you can choose how many factors to keep.
Which predictors matter most, from VIP: values above 1 mark the predictors that contribute most to the projection.
A usable prediction equation — intercept and coefficient for every predictor in the variables' original units — even when the predictors are too collinear for ordinary regression.
How predictors and dependents relate to each latent factor, through the X and Y weights and loadings.

When to use it

Use it when

Many correlated rating items (brand attributes, satisfaction drivers) used to explain one outcome such as purchase intent or overall satisfaction — classic key-driver analysis.
More predictors than you can comfortably fit by least squares, or fewer cases than predictors.
Several related outcomes predicted jointly from the same predictor set.

Use something else when

You need significance tests or confidence intervals for each coefficient: the panel reports none; use linear regression when the predictors are not badly collinear.
You want to choose a small subset of predictors automatically: use automatic linear modeling or predictor selection.
The goal is to summarise predictors without any outcome: that is factor or principal components analysis.
Survey weights must be applied: PLS here is unweighted and the panel says so.

How to run it in Crosstabs

01
Open your data
Open a .sav, CSV or Excel file in the workspace and switch to Statistics mode.
02
Choose PLS regression
In the Regression group, pick Partial least squares regression.
03
Pick the dependents
Tick one or more Scale dependents and/or Categorical dependents. A categorical dependent is coded with one indicator per category except the last.
04
Pick the predictors
Tick categorical predictors under Factors (each category becomes its own indicator) and scale predictors under Covariates.
05
Set the factors and run
Set Number of latent factors (5 by default) and click Run. Cases with a missing value on any selected variable are excluded listwise.

Run Partial Least Squares Regression in your browser

Worked example

Computed by the Crosstabs engine from the synthetic data below when this page was built — not typed in.

Data

60 synthetic shoppers rate a brand on six 0–10 attributes. Quality, reliability and durability share one hidden “product” impression and design and style share a “look” impression, so the predictors are strongly correlated; price_fair is pure noise. Purchase intent depends mostly on the product impression. Fixed seed; Number of latent factors set to 3.

Proportion of Variance Explained

Latent Factors	X Variance	Cumulative X Variance	Y Variance	Cumulative Y Variance (R-square)	Adjusted R-square
1	0.468	0.468	0.806	0.806	0.803
2	0.243	0.711	0.009	0.815	0.809
3	0.056	0.767	0.014	0.829	0.820

60 cases used; all variables are centred and standardised before extraction.

Variable Importance in the Projection

Variable	Latent factor 1	Latent factor 2	Latent factor 3
quality	1.297	1.290	1.282
reliability	1.283	1.277	1.272
durability	1.326	1.324	1.313
design	0.790	0.786	0.828
style	0.535	0.577	0.572
price_fair	0.037	0.039	0.046

Cumulative VIP through each latent factor; values above 1 mark the more important predictors.

Parameters

Independent Variables	intent
(Constant)	−3.4006
quality	0.3205
reliability	0.1945
durability	0.5589
design	0.4563
style	−0.1408
price_fair	−0.0076

Coefficients are in the original units of the variables.

How to read it

The first latent factor alone explains 80.6% of the variance in intent; with 3 factors the cumulative R² is 0.829 (adjusted 0.820) while 76.7% of the predictor variance is used. The later factors add only 2.3 percentage points, so one factor already carries the signal. VIP above 1 marks quality, reliability, durability as the key predictors; the weakest is price_fair (VIP 0.046), the noise attribute in this file. Because PLS extracts a few shared factors, it can use all six correlated ratings at once, where ordinary least squares would struggle with their collinearity.

Output tables explained

Proportion of Variance Explained: For each latent factor: X variance, cumulative X variance, Y variance, cumulative Y variance (R-square) and adjusted R-square. Stop adding factors where cumulative R² levels off; the footnote says if fewer factors could be extracted than requested.
Parameters: The constant and one coefficient per predictor column for each dependent, in the original units of the variables, using all extracted factors.
Variable Importance in the Projection: Cumulative VIP of every predictor through each latent factor, with a bar chart for the final factor. Values above 1 mark the more important predictors; values well below 1 are candidates to drop.
Weights: X weights W for the predictors and Y weights C for the dependents, each column of unit length; they show how each factor is built.
Loadings: X loadings X't/(t't) and Y loadings Y'u/(u'u) in standardised coordinates. Case scores, predictions and residuals can be downloaded as CSV.

Assumptions and notes

Relations between predictors, factors and dependents are linear; PLS handles collinearity, not curvature.
Every variable is centred and scaled to unit standard deviation before extraction, so results do not depend on measurement units; constant columns are dropped.
The number of factors is limited by the rank of the predictors and by N − 1; the panel extracts fewer than requested when it must.
The SPSS PLS algorithm defines no case weights, so results are unweighted; cases with missing values are excluded listwise.

How the results are validated

PLS is checked against two references on a 120-case file in three designs (one scale dependent with a factor and four covariates; two scale dependents; a categorical plus a scale dependent): an independent NumPy transcription of the SPSS algorithm, which matches weights, loadings, scores, VIP, standardised and raw coefficients and predictions to 1e-9 and the variance-explained and adjusted R² columns to 1e-12; and scikit-learn PLSRegression with scaling, which matches coefficients, intercepts, predictions, X weights, Y weights, X loadings and X scores (up to the sign of each factor) to 1e-10 with one dependent and 1e-6 with several. Validation evidence

Questions

How many latent factors should I keep?

Look at the cumulative Y variance (R-square) and adjusted R-square: keep adding factors while they rise noticeably and stop where they level off. Extra factors mostly fit noise in the predictors.

How is PLS different from principal components regression?

Principal components are built to explain the predictors only; PLS factors are built to explain the covariance between predictors and outcomes, so the first PLS factor usually predicts the outcome far better than the first principal component.

What does a VIP above 1 mean?

VIP summarises a predictor's contribution across the factors, weighted by how much Y variance each factor explains; its squares average to 1 over predictors, so values above 1 are above-average contributors.

Can I use a categorical outcome?

Yes. Tick it under Categorical dependents; it is coded with one indicator per category except the last (reference) category, and coefficients are given for each indicator. For classification with probabilities, logistic regression is usually the better tool.

Why are there no p-values?

PLS is a prediction method; the SPSS procedure reports variance explained, VIP and coefficients but no standard errors, and the panel follows that. Judge predictors by VIP and the model by adjusted R².

Is my data uploaded?

No. The PLS model is computed in your browser and the file stays on your device.

Run Partial Least Squares Regression on your own data

Open a .sav, CSV or Excel file. Everything runs in your browser; the file never leaves your device.

Open the workspace

SPSS is a trademark of IBM Corp. Crosstabs is not affiliated with or endorsed by IBM.