AnalyticBridge

A Data Science Central Community

Logistic regression intercept term not significant

presently m building a short stay model for inpatient claims....the dependent variables for the same is defined as the claims for which we recovered amount(that claim can be a waste/fraud claim) by conducting manual audit last year.

Though C stats and other stats are good for the model but the problem starts when i look onto significance of intercept term which is as below:

Parameter	DF	Estimate	Standard	Wald	Pr > ChiSq
			Error	Chi-Square
Intercept	1	-0.0165	0.0728	0.0514	0.8206

M not sure what to do....

can any one please suggest me..what can be done...i understand ..it means that the average value of the dependent variable when all the other independent variables are equal to zero.....but should i include intercept while scoring my model or what else.....?

Thanks in advance..!!!!

Regards

Amitesh

0 members like this

▶ Reply to This

Replies to This Discussion

Permalink Reply by Ralph Winters on September 19, 2014 at 6:49pm

You should always include the intercept. The intercept is just an error term and generally means nothing. All the model is saying is that the intercept happens to be close to zero, but that's it. You can safely ignore the significance

▶ Reply

Permalink Reply by Sagar Diwakar Uparkar on September 23, 2014 at 10:05am

Hi Amitesh,

Good evening. In my opinion it is everyone's own judgement. In this case, try model with and without intercept. The model which will predict more accurate, select that one. It is not mandatory to include intercept.

Warm regards,
Sagar.

▶ Reply

Permalink Reply by Yuecai Zhu on October 2, 2014 at 3:03pm

In my opinion, intercept means unknown factors and error when estimating coefficients in the linear phase . This may be caused by the noise in your training data set if it is mainly error. I think you can try to clean the data and try to add more factors.

▶ Reply

Permalink Reply by Roque Graciani on October 5, 2014 at 9:55am

I may be late to providing a course of action, but I agree with Sagar, cross-validation is probably the best approach, build your model with about 90 percent of your observations and use the model to predict the other 10 percent with and without the intercept, use the model that predicts most accurately.

AnalyticBridge

Logistic regression intercept term not significant

Replies to This Discussion

Resources

On Data Science Central