and

Version 1.3 is only a slight modification of version 1.2 but is provided here because newer PC’s were having difficulty running the old version. Small wonder: version 1.2 was compiled in 1993 for DOS machines. Version 1.3 makes use of extended and expanded memory of Windows and hence is free of the conventional memory restriction that limited the size of data sets that could be handled by version 1.2 (and that in some cases prevented the program from running at all on Windows machines). Version 1.3 is compiled to handle up to 10,000 cases. This value can be increased if necessary by modifying the source code and recompiling (instructions given below). The only other difference is that lambda has been rescaled in version 1.3 to match that now used in the most recent version of the univariate program, GLMS. B) Outline of method This section assumes some prior familiarity with univariate regression using cubic splines. The univariate program GLMS is available from D.S. at www.zoology.ubc.ca/~schluter/splines.html. Example applications of the univariate method are given in Schluter (1988). The present method is based on Schluter and Nychka (1994) which should be consulted for details. The present program estimates fitness surfaces (and other regression surfaces) using th