www.cmm.bristol.ac.uk/forum

Hello everyone,,,
I have collected three waves of data (2007, 2009, and 2011) on the number of minutes that people report walking for transport in the previous 7 days. The sample-size at each wave comprised 200 neighbourhoods and 11,000, 7900, and 6900 respondents in 2007, 2009 and 2011 respectively. Hence I have a three-level repeated measures dataset (neighbourhoods, individuals, time).

The variable at each wave has an excessive number of cases with zero-values (~60%), reflecting the fact that most people didn't walk for transport during the survey reference period; the rest of the cases have non-zero values that range from 1-840 minutes. Hence, the non-zero data arise from a continuous distribution, and are not independent counts.

Tooze et al (Statisitical Methods in Medical Research, 2002;11:341-355) propose a model for repeated measures data with clumping at zero, using a mixed effects mixed distribution model with correlated random effects. The model contains components to model the probability of a non-zero value and the mean of non-zero values, allowing for repeated measurments using random effects and allowing for correlation between the two components. They used the MIXCORR macro in SAS PROC NLMIXED.

Can this type of model be estimated using "runmlwin"?

Many thanks in advance. Regards.

Dear Gujarish,

No you cannot fit this specific model in MLwiN.

For these data, what you could do in MLwiN is fit the analyses in two parts.

First, fit a multilevel model for whether peopled walked at all (0 steps vs 1+ steps)

Then for the subset who had 1+ steps you could fit a multilevel model for the number of steps taken

Best wishes

George

Dear George,

thank you very much for the brilliant LEMMA platform and the resources provided around it. I've been learning about multilevel models and how to apply them in R2MLwiN to use them in my dissertation.
I would like to ask a question related to correlated random effects, so I decided to add it as a reply to Gujarish's initial question.
In my dissertation, I study the provision of social support in personal networks with a rather complex data structure. Essentially, I have respondents' contacts (level 1) being nested in respondents' personal networks (level 2) being nested in geographical areas (level 3). So far, this follows the hierarchical three-level models described in LEMMA and the MLwiN documentation. Additionally to these three levels, I want to study the effects of the respondents' contacts' geographical location (level 2b) leading to a cross-classification of personal networks and contacts' location. Furthermore, as the locations at level 3 and level 2b are the same, their random effects are not independent. Leyland and Naess (https://rss.onlinelibrary.wiley.com/doi ... 08.00581.x) suggest for a similar case a combination of cross-classification and multiple membership model - a correlated cross-classified model.

From the previous discussions on this forum, it seems like these kinds of models are not possible in MLwiN. Leyland and Naess use WinBugs which, I believe, MLwiN also bases on. Is it therefore possible to call WinBugs to run these correlated cross-classified models from MLwiN or R2MLwiN?

Thank you in advance for taking the time.

Best wishes,
Paul