Ydiagnostics {expandFunctions} | R Documentation |
This function presents diagnostic plots of estimate Yhat and response Y.
Ydiagnostics(Y, Yhat, ...)
Y |
R object representing response, coercible to a vector. |
Yhat |
R object representing estimate, coercible to a vector. The length of Y and Yhat must be equal. |
... |
Options for |
The plots shown are:
Y vs Yhat. Under a perfect noise-free fitting, this would be a straight line with the points lined up on the red line, and the correlation wpuld be 1.0000.
Y, Yhat and Y-Yhat (residual) time domain plots. The time steps are in samples.
These show the ACF for the original Y, the residual, and |residual|. The latter helps identify nonlinearity in the residual.
Invisibly returns TRUE; this routine is only used for its graphical side effects described in Details.
# The order here looks backwards, but is chosen to # simulate a typical pair - Yhat will normally have # a smaller range than Y. set.seed(2) nObs <- 100 # Number of observations x <- stats::filter(rnorm(nObs),c(-0.99), method="recursive") x <- x + (x^2) # Nonlinear component myLags <- 0:2 X <- eTrim(eLag(x,myLags)) Y <- X[,+1,drop=FALSE] X <- X[,-1,drop=FALSE] lmObj <- lm(Y ~ X) Yhat <- predict(lmObj) Ydiagnostics(Y,Yhat)