PLS Predict Scaling and Y-Values. Scaling always applied to Y-Values.

[DavidMartinOnGitHub]

The predict function of the PLS Class optionally scales the X values but always reverse scaling on the Y-Values. Shouldn't the Y scaling be embedded within an if(this.scale) block?

Code as-is:

predict(dataset) {
    let X = Matrix.checkMatrix(dataset);
    if (this.scale) {
      X = X.subRowVector(this.meanX).divRowVector(this.stdDevX);
    }
    let Y = X.mmul(this.PBQ);
    Y = Y.mulRowVector(this.stdDevY).addRowVector(this.meanY);
    return Y;
  }

Proposed fix:

predict(dataset) {
    let X = Matrix.checkMatrix(dataset);
    if (this.scale) {
      X = X.subRowVector(this.meanX).divRowVector(this.stdDevX);
    }
    let Y = X.mmul(this.PBQ);
    if (this.scale) {
       Y = Y.mulRowVector(this.stdDevY).addRowVector(this.meanY);
    }
    return Y;
  }

[josoriom]

Indeed, the code reverses the scaling of Y, but I think it is acceptable because when I make predictions on new data, I expect the response to be in the same scale as my input data. However, the PLS class provides a toJSON method that contains the necessary information to rescale Y if needed.