A study was made to test and compare the behavior of a standard non-linear regression program (BMDP3R) in fitting data from six classical least-squares problems. The use of three program control parameters is discussed and four measures of regression failure are utilized to give a quantitative reference of success. Recommendations are given to aid the user of packaged programs in the parameter estimation of non-linear regression models.
Download full-text PDF |
Source |
---|---|
http://dx.doi.org/10.1016/0010-468x(79)90066-7 | DOI Listing |
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!