Prediction of new data points with the Pearson kernel.

[haziqj]

Suppose we fit an I-prior model mod <- iprior(y~x, data), and we use the predict function to obtain the predicted value at x[1]. Why doesn't this value match with fitted(mod)[1]?

Suspect that there is a wrong implementation of the Pearson kernel code. The Pearson kernel is h(x_i, x_j) = I(x_i = x_j) / p_i - 1. When a new data point x_k is introduced which has the same value as x_i, does p_i change as well, or does it remain the same? p_i is the proportion of the data which has the same value as x_i.

[Wicher2]

Hi haziq, probably the most logical is not to change the value of p_i.
Wicher

[haziqj]

Ah ok, then this needs fixing. At the moment the value of p_i does change.