
 PA 765: Multiple Regression 
  Multiple regression shares all the assumptions of correlation: linearity of relationships, the same level of relationship throughout the range of the independent variable ("homoscedasticity"), interval or nearinterval data, absence of outliers, and data whose range is not truncated. 
  Cubic regression splines operate similar to local polynomial regression, but a constraint is imposed that the regression line in a given bin must join to the start of the regression line in the next bin, thereby avoiding discontinuities in the curve, albeit by increasing error a bit. 
  Local regression fits a regression surface not for all the data points as in traditional regression, but for the data points in a "neighborhood." Researchers determine the "smoothing parameter," which is a specified percentage of the sample size, and neighborhoods are the points within the corresponding radius. 
 www2.chass.ncsu.edu /garson/pa765/regress.htm (19061 words) 
