Full Regression Model:

Measles Immunization


Return to Multi-way Page 3

To run the full regression with the two interaction variables (MEAS_AGE, and LOWED_HT), just follow these steps.  Then scrutinize your results to see if you think the model is appropriate:

1.  Open keast4j.sav
2.  Click on Statistics, Regression, Linear.
3.  For the Dependent variable, enter waz using the arrow.
4.  For the Independent variables, enter age, agesq, hmeasyn, dlowedn, htresp, meas_age, and lowedn_ht.
5.  Click on OK.

 

Do your results look like these?

 

wpe9.jpg (12510 bytes)

wpeB.jpg (13602 bytes)

wpeC.jpg (25484 bytes)

INTERPRETATION:

You are likely already realizing that there is something likely to be wrong with this analysis, since you have a negative effect from measles according to the B coefficient (-0.253).  As we have seen in an earlier section, there is certainly some confounding of age and measles immunization since those younger children do not get immunization and they are also likely to be better nourished. Here, it appears that the interaction variables are both non-significant which would indicate that you should try running the model again without them. Because measles and age is thought to have a strong likelihood of interacting, we will keep it in the model even though it is does not have a significant p-value.  Really, the next best step to see what is happening with measles and nutrition status, lets look at these variables in a model that is stratified by age.

 

But first, here is a look at the results of the full model without the interaction the interaction term for education and height.

wpeD.jpg (11660 bytes)

wpeE.jpg (13897 bytes)

wpeF.jpg (23742 bytes)

Return to Top