I am planning to conduct a research on " Socio-economic determinants and inequalities in maternal and child health care service utilization". I would like to model multilevel analysis. However, am not sure how much is the minimum level of clusters needed to use multilevel models.Therefore, I would be very happy and grateful if anybody provides me an information. in addition how can I collect health care facility and other community level data's? in addition to household level data.
Dear Asmamaw Atnafu Ayalneh,
The answer depends partly on sample size and power at each level. For a quick introduction see:
Peterson MF, Arregle J-L, Martin X. 2012. Multilevel models in international business research. Journal of International Business Studies 43(5): 451–457.
For an in-depth review I recommend:
Hox JJ. 2010. Multilevel analysis: Techniques and applications (2nd ed.). Routledge: New York.
Just to elaborate on Xavier's point. The power is different at each level so it also depends on the questions you are asking. N at level 2 may not be as important if you are most interested in fixed effects at level 1 (for instance), though one should be aware that low level 2 usually results in underestimated SEs.
I think you are looking at this from the wrong end of the telescope! That is: do you have research questions that requires multilevel modeling to provide a good answer?
Multilevel modeling is needed when your data has a complex structure and this type of modeling is required either because of technical requirements - eg dependency or some types of missingness - or because of substantive questions eg contextual questions involving micro and macro levels simultaneously. These structures can be quite complex and this may be helful to you in thinking about structures
Multilevel modeling the background - structures and concepts
http://www.bristol.ac.uk/cmm/software/mlwin/mlwin-resources.html#mlmbkgrd
and these are some papers in the health field
https://www.researchgate.net/publication/21295937_Multilevel_Assessment_of_Immunisation_Uptake_as_a_Performance_Measure_in_General_Practice
https://www.researchgate.net/publication/13765545_Context_Composition_and_Heterogeneity_Using_Multilevel_Models_in_Health_Research
Article Context, Composition and Heterogeneity: Using Multilevel Mod...
Article Multilevel Assessment of Immunisation Uptake as a Performanc...
I thank You all,
Dear Kelvyn Jones, I understood what you mentioned about the research question and the need of multilevel modeling. However, when I saw several researches regarding health care services utilization they are considering only individual factors and lack other contextual factors ( area level and community factors). That is why I wanted to look for the effect of both these factors on maternal and child health care service utilization by using multilevel modeling.
Number of clusters depend on sample size and variation in the reposes given by the respondents
Just to add you may need a cross-classified that is more complicated and realistic than a strictly hierarchical model. model so individuals are nested in structures that are crossed eg individuals nested within primary care facility - secondary care and residential zone - they may attend more than one care facility so that would need a multiple membership model. This is very likely if you are modeling repeated measures over time as people move. When analyzing these sort of structures an effective way of deciding on how many clusters is by simulation and this software was designed to do that:
http://www.bristol.ac.uk/cmm/software/mlpowsim/
- Badges
- Science topic
- Similar topics
- Medicine
- Counseling Psychology