+2008-08-17 Jason H Stover <jhs@math.gcsu.edu>
+
+ * linreg.c (post_sweep_computations): New function to re-use the
+ re-usable. New function cache_init() to re-use the re-usable.
+ pspp_linreg_with_cov() now estimates parameters via reg_sweep().
+
+2008-08-16 Jason H Stover <jhs@math.gcsu.edu>
+
+ * linreg.c (pspp_linreg_with_cov): New function to estimate
+ parameters directly from covariance matrix instead of using the
+ entire data set.
+
+ * linreg.c (rearrange_covariance_matrix): New function to ensure
+ the columns of the covariance matrix correspond to the variables
+ in the model.
+
+2008-07-24 John Darrington <john@darrington.wattle.id.au>
+
+ * merge.c merge.h sort.c sort.h: Removed the value_cnt associated
+ with case ordering. This ensures that the casereader returned by
+ sort_execute now has the same dimensions as the casereader passed
+ in.
+
+2008-07-24 Jason H Stover <jhs@math.gcsu.edu>
+
+ * covariance-matrix.c (covariance_pass_two): Dropped weight
+ argument. Corrected update of the entries of the covariance
+ matrix.
+
+ * covariance-matrix.c (column_iterate): Dropped weight
+ argument. Corrected update of the entries of the covariance
+ matrix.
+
+ * covariance-matrix.c (covariance_update_categorical_numeric):
+ Dropped weight argument. Corrected update of the entries of the
+ covariance matrix.
+
+2008-07-17 Jason H Stover <jhs@math.gcsu.edu>
+
+ * covariance-matrix.c (covariance_update_categorical_numeric):
+ Altered to match new code in covariance_pass_two();
+
+ * covariance-matrix.h: Fix declaration of covariance_pass_two.
+
+ * covariance-matrix.c (covariance_pass_two): Moved numeric and
+ categorical scopes of V2 inside for loop. If V1 is not
+ categorical, but V2 is, recurse with the order of the variables
+ reversed to use the code in the previous "if" scope.
+
+2008-07-16 Jason Stover <jhs@math.gcsu.edu>
+
+ * covariance-matrix.c (covariance_pass_two): Renamed
+ covariance_pass_one. Fixed update of covariance matrix when both
+ variables are categorical.
+ (column_iterate): New function.
+
+2008-07-15 Jason Stover <jhs@math.gcsu.edu>
+
+ * covariance-matrix.c (covariance_pass_one): New file, new function.
+
+2008-06-21 Jason Stover <jhs@math.gcsu.edu>
+
+ * linreg.c (pspp_linreg): Accept a struct design_matrix. Use new
+ accessor functions pspp_coeff_get_mean, pspp_coeff_get_sd. New
+ functions pspp_linreg_get_indep_variable_mean,
+ pspp_linreg_set_indep_variable_mean,
+ pspp_linreg_get_indep_variable_sd,
+ pspp_linreg_set_indep_variable_sd. Altered pspp_linreg_get_coeff
+ to use accessor function pspp_coeff_var_to_coeff.
+
+ * coefficient.c: New functions pspp_coeff_get_mean,
+ pspp_coeff_set_mean, pspp_coeff_get_sd, pspp_coeff_set_sd,
+ pspp_coeff_var_to_coeff. Added doubles to hold mean and standard
+ deviation in struct varinfo.
+
+2008-06-14 Jason Stover <jhs@math.gcsu.edu>
+
+ * linreg/: moved linreg.[ch] to src/math.
+
+2008-05-15 Ben Pfaff <blp@gnu.org>
+
+ Patch #6512.
+
+ * moments.c (calc_moments): Use gsl_finite instead of finite, as a
+ stopgap measure for portability until appropriate gnulib modules
+ are available.
+
+2008-03-10 Jason Stover <jhs@math.gcsu.edu>
+
+ * coefficient.c (pspp_linreg_get_coeff): Removed use of
+ coefficient 0 as intercept.
+
+2008-03-01 Jason Stover <jhs@math.gcsu.edu>
+
+ * coefficient.c (pspp_coeff_init): Ensure first arg is not a null
+ pointer.
+
+ * coefficient.c (pspp_linreg_get_coeff): Make sure we don't return
+ a result beyond the last coefficient, or start with a coefficient
+ beyond the last one if there is only one.
+