Comments - Simple technique to improve poor predictive models - AnalyticBridge2018-11-15T01:40:43Zhttps://www.analyticbridge.datasciencecentral.com/profiles/comment/feed?attachedTo=2004291%3ABlogPost%3A220409&xn_auth=noVincent. I'm not sure what t…tag:www.analyticbridge.datasciencecentral.com,2012-11-03:2004291:Comment:2216562012-11-03T14:16:58.835ZRalph Wintershttps://www.analyticbridge.datasciencecentral.com/profile/RalphWinters
<p>Vincent. I'm not sure what the statistical basis behind this method is. Would you advocate jumbling all of the coefficients of a bad linear regression model until it led to an improved best r-square?</p>
<p>-Ralph Winters</p>
<p></p>
<p>Vincent. I'm not sure what the statistical basis behind this method is. Would you advocate jumbling all of the coefficients of a bad linear regression model until it led to an improved best r-square?</p>
<p>-Ralph Winters</p>
<p></p> Very interesting concept - I'…tag:www.analyticbridge.datasciencecentral.com,2012-11-01:2004291:Comment:2206302012-11-01T09:16:09.461ZEli Y. Klinghttps://www.analyticbridge.datasciencecentral.com/profile/EliYKling
Very interesting concept - I'll have to try it.<br />
One thing you might sharpen in the above text is that you are discussing sequnces of predictions (or modeled sequnces).<br />
I am not sure cross validation is that straight forward to implement.<br />
Just a thought - using this general idea in non time series situations: do a quick chaid on the prediction erorrs on a test or validation set to evaluate the conditional bias drivers (if they exist)
Very interesting concept - I'll have to try it.<br />
One thing you might sharpen in the above text is that you are discussing sequnces of predictions (or modeled sequnces).<br />
I am not sure cross validation is that straight forward to implement.<br />
Just a thought - using this general idea in non time series situations: do a quick chaid on the prediction erorrs on a test or validation set to evaluate the conditional bias drivers (if they exist)