MATLAB: What is the difference between RESID, COMPARE and PE functions in System Identification Toolbox 7.2.1 (R2008b)

Question

I would like to know the difference between RESID, COMPARE and PE functions in System Identification Toolbox.

Answer 1

RESID and PE functions compute 1-step ahead prediction errors (M=1). For errors

corresponding to arbitrary horizons, you may compute prediction results

yp = predict(model, data, M),

where M is prediction horizon. You could also use COMPARE function too whose first output argument returns the predicted output:

Y = compare(data,model,M)
yp = Y{1};

Note that calling "COMPARE(data, model)" means prediction horizon M is Inf. You can also compute prediction error as difference between measured and predicted outputs as follows:

e = data;
e.y = data.y - yp.y;