Hi community fellows, Is anyone familiar with the definition of stderr (standard error) of predicted values (coefficients and dependent variable) in GWR? For OLS (ordinary least square), given a set of data (like n samples), one can get only one set of coefficients and n errors ??residuals), consequently, the stderr is the standard deviation of these n errors. However, for GWR, if it is based on the same idea or what? The confusing part for me is that as GWR gives weights to each sample, when doing least square fitting, it means that it gives weights to each residual and finally find the optimal coefficients to minimize the summation of these squared weighted-residual. Now, problem comes out with the calculation of stderr is that if the weighted residual should be used to estimate the stderr for predicted dependent variables or the way that used in OLS is still hold here? Anyone giving the direct answer or reference would be appreciated. Gavin
... View more