1
$\begingroup$

I have hierarchical data of individuals nested into families. For each individual, I have independent variables such as age, gender, education, and familiarity with product. For each family unit, I also have covariates such as household income, purchase behavior, and distance to retail centers.

The dependent satisfaction measure is only recorded at the family level. More specifically, satisfaction is asked of a head-of-household respondent, who ideally represents the household. While satisfaction is measured on a 5-point scale, we typically re-express it as dichotomous (top 2 box).

I would like to take into consideration the individual-level effects as well as the family-level effects in modeling product satisfaction propensity. Is it appropriate to explore multilevel modeling when the outcome is only measured at the second level? If not, is there a different approach I should be following?

$\endgroup$
3
$\begingroup$

A nice paper about this is the following:

Basically the approach that they outline involves computing adjusted group means on the predictor variables and then regressing the outcome on the adjusted group means. The adjusted group mean for each group is the best linear unbiased predictor (BLUP) of the predictor variable for that group. You can compute those using equations given in the paper or, if you're using R, using the lme4 package and its coef() function.

$\endgroup$
  • $\begingroup$ To calculate the adjusted group means on a predictor X, is it sufficient to simply use lm(X~ 0 + groupID) $\endgroup$ – Amw 5G Sep 1 '15 at 22:13
  • $\begingroup$ @Amw5G No, that would just yield the simple means, not the adjusted (shrunk) means. But you could use something like mod <- lmer(X ~ 1 + (1|groupID)); coef(mod) $\endgroup$ – Jake Westfall Sep 1 '15 at 23:42
  • $\begingroup$ Ah, I understand now. Much obliged @Jake Westfall $\endgroup$ – Amw 5G Sep 2 '15 at 0:23
1
$\begingroup$

This is certainly doable. I think that the best approach would be to use a multilevel SEM package (e.g., MPlus, Stata gsem, or R lavaan) that allows you to specify which level your variables are at. Note that with a level 2 outcome, all regression paths will be from L2 (latent) aggregates to the outcome.

$\endgroup$
  • $\begingroup$ Thank you for your reply @erik r.; I found your suggestions of example software particularly helpful as I'm using R, but have colleagues with SEM experience in MPlus. $\endgroup$ – Amw 5G Sep 1 '15 at 16:43
  • $\begingroup$ The really nice property of a program like Mplus is that you can specify your variables as having both L1 and L2 variance (if appropriate). Such variables are treated as latent within and between components, and used for estimating the regression paths to your dependent variable. No extra calculation of the adjusted group means is necessary, and many would argue that keeping everything latent is the more appropriate approach (certainly it is more parsimonious). That being said, the BLUP approach identified by @Jake Westfall will do the trick, especially if you don't have access to Mplus. $\endgroup$ – Erik Ruzek Sep 2 '15 at 17:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.