Re: weighting

Posted by Art Kendall on
URL: http://spssx-discussion.165.s1.nabble.com/weighting-tp5650892p5651735.html

The regression coefficients are the weights that variables have when summing them.
predicted Y = weight1*intercept + weight2*IV1 + weight3*IV2 ...

It is unclear whether you want a) to create an equation and find the weights  or  b) whether you want  the weight for age to be .30 and the weight for income to be .27 then find the weights for the other variables.

In order for list members to give clearer advice we would need to have a clearer question.   Please describe the problem including answers to the questions I posted earlier.

Are you interested in predicting churning for other cases?  Are you interested in distinguishing churners from non-churners?

On 4/19/2012 7:32 AM, Rajeshms wrote:

Oh My god....................ha ha,I am sorry Art.

Ok Here its is its not a homework. I am a fresh statistician and part of job.

I want to do linear regression and I have some weights on each IVs.So how can I do.Say for instance I need 30% weights of Age and 27% on income.....which are IVs. So this was what I understood. Thank you very much for your time.

regards

rajesh 

On Thu, Apr 19, 2012 at 4:43 PM, Art Kendall <[hidden email]> wrote:

It would be necessary to have more detail so that list members can try to help.

You say you are fitting a model.  Are you asking how to fit a model?

What is the model for? Do you want to gain more understanding of churning or are you just interested in developing a model to predict future cases?

What is your definition of churning?  Are you stuck with such a coarse measurement of your DV?

What is a case in your data?  How did you select the cases? How many do you have?

Are all of your predictors continuous?

What is the context? Is this homework? Part of a job? etc.

It seems that the churn rate would simply be the proportion of cases that had churn.  Please explain. 

Google incidence vs prevalence.  Which are you interested in?

Art Kendall
Social Research Consultants

On 4/19/2012 1:08 AM, Rajeshms wrote:

Hi All,

I had a query in fitting a model where we are interested in finding the churn rate.

I am fitting a model with DV as Churn (yes/no) and IV as age , tenure, income and etc.

But now we need to weight each IV separately and fit a model........I don know how to do this using spss, I hope anyone will help me to solve this.........thanks in advance.this will be helpful to my career.

--

Rajesh M S

--

Rajesh M S

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD