Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

how to calculate total estimated value for glm model in sas studio and enterprise miner

Reply
Occasional Contributor
Posts: 19

how to calculate total estimated value for glm model in sas studio and enterprise miner

Hi,

I am using Genmode procedure with Negative binomial model to model total claims for a motor insurance data. 

below is the code:

proc genmod data=libish.simulatedwithlog;

class premiumclass age zone;

model no=  zone freq   /dist=negbin offset=logduration type3;

output out=libish.dataset2 predicted=nbclaimestimate;

run;

 

the Output statement here creating a new variable called nbclaimestimate , Is this estimated No. of claims calculated by model? I want to calculate total number of estimated claims , can i use these estimated values to show my result

 

or i need  to calculate this manually by using intercept and covariates exponential like:

 exp(Intercept) * exp(Zone) * exp(freq).........

 

i have one more doubt: i have some example using exp(Intercept) * exp(Zone) +exp(freq)

what criteria decides additive effects or multiplicative effect? sorry i am a learner and  new in this field , please help  a bit to understand this.

Occasional Contributor
Posts: 19

How to calculate Avg estimated value in GLM

Hi, 

 

 

I am using Genmode procedure with Negative binomial model to model total claims for a motor insurance data. 

below is the code:

proc genmod data=libish.simulatedwithlog;

class premiumclass age zone;

model no=  zone freq   /dist=negbin offset=logduration type3;

output out=libish.dataset2 predicted=nbclaimestimate;

run;

 

the Output statement here creating a new variable called nbclaimestimate , Is this estimated No. of claims calculated by model? I want to calculate total number of estimated claims , can i use these estimated values to show my result

 

or i need  to calculate this manually by using intercept and covariates exponential like:

 exp(Intercept) * exp(Zone) * exp(freq).........

 

i have one more doubt: i have some example using exp(Intercept) * exp(Zone) +exp(freq)

what criteria decides additive effects or multiplicative effect? sorry i am a learner and  new in this field , please help  a bit to understand this.

Trusted Advisor
Posts: 1,253

Re: How to calculate Avg estimated value in GLM

Hi,

 

Youve posted your question in the SAS Visual Analytics community. You will probably get a response your if you move the question (or recreate it) in the SAS Statistical Procedures community https://communities.sas.com/t5/SAS-Statistical-Procedures/bd-p/statistical_procedures

 

Kind Regards,

Michelle

Super User
Posts: 17,907

Re: how to calculate total estimated value for glm model in sas studio and enterprise miner


proc genmod data=libish.simulatedwithlog;

class premiumclass age zone;

model no=  zone freq   /dist=negbin offset=logduration type3;

output out=libish.dataset2 predicted=nbclaimestimate;

run;

 

the Output statement here creating a new variable called nbclaimestimate , Is this estimated No. of claims calculated by model? I want to calculate total number of estimated claims , can i use these estimated values to show my result

 

 


Yes, nbclaimEstimate is the predicted value of the number of claims. 

You can use these estimated values, as long as you're confident in your model. 

 

If you're unsure of your analysis method, your best bet is to review the statistical methodologies behind the model. Why did you pick Genmod in the first place?  How did you decide on the dis/offset?

That will help you interpret your results and understand why the model is the way it is in terms of components.

Occasional Contributor
Posts: 19

Re: how to calculate total estimated value for glm model in sas studio and enterprise miner

Thanks Reeza

Super User
Posts: 9,687

Re: how to calculate total estimated value for glm model in sas studio and enterprise miner

This could give you a better explain .

http://support.sas.com/kb/24/188.html 

Ask a Question
Discussion stats
  • 5 replies
  • 288 views
  • 1 like
  • 4 in conversation