Question: If a regression spline were fit to these data, what order of polynomial might be used? Similarly, where should the knot(s) be placed? Fit the regression spline and compare the smooth with the smooths produced in Problems II and I.
Problems I: Remembering that a running line smooth can be fooled by anomalous data points, use the Minitab macro RLINSMOL with 25% trimming to produce a robust running line smooth, again using k = 15, 25, and 35. Compare the results with the smooths obtained for Problem II Is there evidence of instability, as in the example, or does the smooth appear to be well determined?
Problem II: Use either the Minitab macro RLINESMO or other software to produce the running line smooth for k = 15, 25, and 35, and compare the results.