The Kernel Density Estimate

Click the Run button below when the applet has completed loading.

This applet illustrates kernel density estimates (KDEs). The applet allows the user to select one of several populations from which to generate samples. Once a sample is generated, the user may use the slider bar to select the smoothing parameter for estimating the density. The true density is overlaid for comparison. The user may also select from one of seven kernel functions.

For each new sample, the window width width is set to the value

       w = IQR(x) / 1.349
       if (w < SDEV(x)) w = SDEV(x)
       width = 0.9 * w / n^(1/5)