reset password
Author Message
alomo
Posts: 70
Posted 19:45 Mar 31, 2010 |

In the article (p.234, figure 5) it is demonstrated that no improvement was achieved by incorporating variance weighting into the Pearson correlation coefficient formula. Quite the reverse, the mean absolute error increased. So, is the variance weighting useful after all?

cysun
Posts: 2935
Posted 12:23 Apr 01, 2010 |

Based on the experimental results, the answer is no. This shows why conducting experiments is so important -- often times some intuitive idea "makes lots of sense", but it may or may not work out in the real world.