Model validation measures

Sweetea — Wed, 16 May 2018 14:55:15 GMT

Hi all,

I have no show predictive modeling built for a clinic using SAS EM. The target is binary (show or no show). I have calculated the following validation measures . I have convinced on all measures outcome but sensitivity is very low (5%).

Is there any ways that would help to increase sensitivity? What factors are associated with sensitivity?

Sensitivity 5.660377358
Specificity 99.66181941
Yeild 1.214689266
PPV 76.74418605
NPV 84.27223334

		Actual
		No Show	Show
Predicted	No Show	33	10	43
Show	550	2947	3497
		583	2957	3540

Sensitivity	5.66037736
Specificity	99.6618194
Yeild	1.21468927
PPV	76.744186
NPV	84.2722333

Please advise

Thanks

Re: Model validation measures

MelodieRush — Fri, 01 Jun 2018 14:42:03 GMT

Changing the cutoff value can improve the sensitivity. Look for information on changing the cutoff value in EM help under the Assess Nodes --> Cutoff Node.

The default cutoff is .50. Lowering that value can often increase sensitivity.

topic Re: Model validation measures in SAS Data Science

Model validation measures

Re: Model validation measures