Sharmistha -
Heteroscedasticity is not just a matter as to existence, but rather degree.
To measure heteroscedasticity, I suppose you could use SPSS, but I do not know modern SPSS. If you can follow the algorithm for the Iterated Reweighted Least Squares (IRLS) method, you will find a good explanation in Carroll and Ruppert(1988), Transformation and Weighting in Regression, Chapman & Hall, Ltd. London, UK.
For a method of measuring the coefficient of heteroscedasticity for a well accepted regression weight format, based on regression diagnostics, see this:
https://www.researchgate.net/publication/263809034_Alternative_to_the_Iterated_Reweighted_Least_Squares_Method_-_Apparent_Heteroscedasticity_and_Linear_Regression_Model_Sampling
This is in accordance with the finite population orientation used here:
https://www.researchgate.net/publication/262972023_HETEROSCEDASTICITY_AND_HOMOSCEDASTICITY,
but could involve time.
For more on measuring the coefficient of heteroscedasticity, see this:
https://www.researchgate.net/publication/263032446_Weighting_in_Regression_for_Use_in_Survey_Methodology
Besides examples shown in a number of papers on my ResearchGate pages,
https://www.researchgate.net/profile/James_Knaub/publications,
here are a couple of interesting uses:
Under 1995 for
http://ww2.amstat.org/sections/srms/Proceedings/
See
Steel, P. and Fay, R.E. (1995), “Variance Estimation for Finite Populations with Imputed
Data,” Proceedings of the Section on Survey Research Methods, Vol. I, American
Statistical Association, pp. 374-379,
and
Sweet, E.M. and Sigman, R.S. (1995), “Evaluation of Model-Assisted Procedures for
Stratifying Skewed Populations Using Auxiliary Data,” Proceedings of the Section on
Survey Research Methods, Vol. I, American Statistical Association, pp. 491-496.
Cheers - Jim
Conference Paper Alternative to the Iterated Reweighted Least Squares Method ...
Article HETEROSCEDASTICITY AND HOMOSCEDASTICITY
Article Weighting in Regression for Use in Survey Methodology
I see that you can start with the diagnostics shown in the attachment that Santam provided.
However, you need to estimate the coefficient of heteroscedasticity by using one or more methods found through links above. Also note that for problematic data quality among the y-values collected for the smallest x-values, you may want to underestimate the coefficient of heteroscedasticity. For highly skewed establishment survey data, I have generally gone to the low end of the reasonable scale (as noted by Ken Brewer), and I therefore have often used the classical ratio estimator for those survey data. Note also that for multiple regression, one may use a preliminary prediction-of-y, rather than x, as the size measure used in the regression weight, along with the coefficient of heteroscedasticity.
Thanks a lot for the materials. They definitely helped a lot
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS