Heteroscedasticity Assumption in Regression

[Neeraj Kaushik]

Assumption of Heteroscedasticity states that Variance of residuals shd be constant against dep var, any of the indep var or time. When some indep var(s) is associated with residual then this problem occurs.

How to check it in SPSS?

Working file: mallcost.sav

Dep var: Adjusted cost of construction

Indep var: Square footage

1) Click on Analyze->Regression-> Linear Regression 

Provide both dep & indep var

2) Click on Save-> Residuals->Standardized

3) Forget the output

4) Click on Graphs-> Legacy Dialogs->Scatter->Simple Scatter

Put Standardized residuals in y-axis, square footage in x-axis and click on OK

 

You'll get the plot as given in Residual plot.docx file

 

The output must be uncorrelated but here see when we move on x-axis there seems a funnel shape of the plot, this shows that when we increase the value of x the corresponding variations in y-axis (signifying residuals) increases; thus the 2 are related and this is called as Heteroscedasticity.

 

Even when there are several indep var, there will be only 1 column for 

the residual as its the tbetweenhe actual & predicted value of Y.

 

 What to do now?

We'll use the Weight Estimation Regression about which I'll write in next post.

 

 

Happy Learning

Neeraj

 

 

 

 

 

[Naveen Dua]

Neeraj Sir,

 Thanks for explaining the concept and procedure to check heteroscedasticity and providing residual plot file. I will try to check the same for my data file and see what is the shape of the plot and  ask for ur guidance. Will be inquisitive to know the way to correct the problem.

Regards

 

--
You received this message because you are subscribed to the Google Groups "DataAnalysis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataanalysistrai...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Naveen Dua]

Sir

(1) why do we save standardized residual in this case and not unstandardized one?

(2) In my case, in multiple regression analysis, there are 5 independent variables and few dummy variables. In the simple scatter plot on X axis, it accepts only one independent variable at a time. so I get 5 diff. scatter plots. Is it the right way ?

From: Neeraj Kaushik <kaushi...@gmail.com>
To:
Sent: Friday, 5 April 2013 6:09 PM
Subject: Heteroscedasticity Assumption in Regression 

[Neeraj Kaushik]

R/Mam

I tried working on unstd residuals and graph is the same.

Actually SPSS gives the option of plotting zresid (Standardized Residual) & zpred (Standardized predictor).

This can be done by Clicking on Plot and putting zresid on y-axis and zpred on x-axis.

Alternatively we can plot unstandardized residuals as well as unstandardized indep var. The plot is the same.

When there are >1 indep var we need to check indivudual var for the possibility of Heteroscedasticity.

I shall write about the Weight estimation Regression soon.

Regards

Neeraj

[Naveendua]

Neeraj sir

Thanks a lot for all ur guidance. 

Regards

Sent from my iPhone

[drsajeeb]

Respected Neeaj Sir,

It's help to all of us. Thank you.

Regards,

Sajeeb Kumar Shrestha

Faculty, Tribhuvan University

Nepal

[KRISHNA KUMAR MISHRA]

thanks sir

--

You received this message because you are subscribed to the Google Groups "DataAnalysis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataanalysistrai...@googlegroups.com.

To post to this group, send email to dataanalys...@googlegroups.com.
Visit this group at http://groups.google.com/group/dataanalysistraining.
For more options, visit https://groups.google.com/d/optout.