topic Re: Average Square Error way to high ?? in SAS Procedures

Average Square Error way to high ??

saraiya0090 — Fri, 22 Mar 2019 12:33:40 GMT

Hello guys,

Currently, I am doing one project on SAS Enterprise Miner, just needed one help while doing modelling I faced an issue of skewness with my continuous variables for which I firstly transformed into log but it didn't work but when changing it into square I got my answer and skewness issue was solved further I did modeling and the result was very extreme, my target variable was estimation of sales basically continuous, and modeling approach was linear regression, but I got validation average square error as 116 so I am confused how to interpret such huge error secondly when the same transformation is done through log the error rate is 0.25 but the continuous variable is not skewed.

Please, your efforts will be appreciated.

Thank You in advance.

Re: Average Square Error way to high ??

Sajid01 — Sun, 27 Jun 2021 17:22:19 GMT

You have squared your continuous variable. That may be the cause.

Re: Average Square Error way to high ??

PaigeMiller — Sun, 27 Jun 2021 18:32:50 GMT

I faced an issue of skewness with my continuous variables ...

Are you talking about skewness of the target (or response) variable? Or skewness of the predictor variables?

... for which I firstly transformed into log but it didn't work

What didn't work? What did you see that makes you think it didn't work? What variable(s) did you transform?

but I got validation average square error as 116 so I am confused how to interpret such huge error

Did you transform the target variable? You really haven't said if you did or did not.

What makes you think that a 116 is "huge"? What are typical values of sales? Maybe the interpretation is that the model didn't fit well; or maybe the interpretation is that you have very large values of sales and a 116 is reasonable. We need a lot more information and context here.