Level 2 Help for IBISREGR

 Input IBIS tabular file containing the data to be fit.

 An output IBIS tabular file containing the NBEST best solutions.

 The column in the input file containing the dependent variable.

 The default is the last column.

 The columns in the input file through which to search (i.e., those
 containing all possible combinations of independent variables).

 The default is all columns except the last.

 The minimum and maximum number of combinations of columns to try.
 The program will start by examining all solutions using COLRANGE(1)
 columns (selected from those specified by the COLS parameter), then
 all solutions using [COLRANGE(1)+1] columns, up through COLRANGE(2)
 columns.

 The default is (2,NCOLS), where NCOLS is the number of columns
 specified in COLS.

 The number of highest-ranked solutions to retain, in order of the
 criteria specified by the CRITERIA parameter.

 An optional eight character heading for the dependent variable.

 Optional eight character headings for the columns containing the
 independent variables.

 Criteria for the best solution(s).  Valid values are:

 RSQUARE:  only the R-squared value will be used as criterion (a higher 
  value is better).

 STDERR:  only the standard error will be used as criterion (a lower value
  is better).

 BOTH (default):  both the R-squared value and the standard error will be
  used as criteria, i.e., a solution is only accepted as better than a
  previous one if both criteria are satisfied.  

 EITHER:  a solution is accepted as better if either the R-squared or the
  standard error criterion is satisfied.

 Note that for BOTH and EITHER, the order in which the solutions are
 encountered will influence the final result if the two do not agree: if
 one solution is best by one of the criteria and a different one by the
 other, then in the case of BOTH the first solution encountered will be
 chosen and in the case of EITHER the second one.

 In general, the R-squared criterion will favor solutions using more
 dependent variables, whereas using more variables will be somewhat
 penalized by the standard deviation.
 

 Print output for each iteration?