Hello all,

My question is about multilevel modeling. I have a data which observed household members food acquisition for one-week. In total this dataset includes 4826 households, 7086 individuals, and 26841 food acquisition events. About % 84 of total food acquisition (more than 22.000 events) was done by household primary shopper who are main responsible person in each household and rest of the acquisition events were done by other household members.

I would like to analyze how household, event, and individual characteristics affect distance of places which people shop. So, I believe that there are some multilevel factors such as household level, (income and surrounding places etc.), individual level (age, gender, employment, primary shopper or not etc.), and event level (place type supermarket, grocery store, restaurant, and whether event was free, etc). Also, in the dataset, some individuals were observed multiple times because they did food shopping multiple times observation week. I would like to ask you whether multilevel regression concept seems appropriate in this context? If so, should I check interclass correlations (ICC) before start to analysis? When I type the command "icc eventid pnum hhnum" (eventid; unique event number, pnum household member id, hhnum; unique household id) I got the following error message:

"multiple observations per target and rater not allowed" r (498)


I also attached a word file describing what my data look like in case you may want to see it. I hope to hear your response and feedback. Thanks in advance.