Search

Multilevel methods have been used for a relatively long time, but only in the last decade have the simple tools for applying them been provided by the most common statistical software packages. Many books and working documents have been published on these models in recent years, with their practical implementation being explained by, for example, Wang (2012) and Givord and Guillerm (2016). Few, however, are suited to beginners. Because this type of analysis can present many obstacles at diffe...

Before conducting any type of multilevel modelling, it is important to first run the model at the individual level, which is what we do in this chapter using a simple logistic regression. A few other models are also run at the individual level, although these will integrate contextual variables. 1. Performing a Logistic Regression at the Individual Level Given the permitted values of the variable to be explained (school, which takes only one of two values: 0 or 1), linear regression is not ap...

Statistical databases increasingly allow for multilevel analyses of both Global North and Global South countries. In some cases, data are collected specifically for this purpose and thus include individual- and meso-level information defined according to the research problem. This can consist, for example, of companies that employ the surveyed individuals (paired family and employer surveys in the North), places of residence (the Migration Dynamics, Urban Integration, and the Environment [EMI...

1. Choosing a Modelling Strategy Before implementing a model, it is important to consider which modelling strategy to use. A key problem in multilevel analysis concerns how to logically integrate the explanatory variables and in what order. Thus, regression can be approached in two opposite ways: one is additive, the other subtractive. The first begins with an empty model and integrates the different explanatory dimensions step by step, while the second begins with a saturated model and succe...

Demography brings together multiple levels of observation. It analyses phenomena on the scale of populations, but also of the individuals they comprise and the events, from birth to death, that punctuate people’s lives. Describing and analysing individual events, either direct or indirect factors in how population structures evolve, thus call for a simultaneous use of both individual and population levels. Demographic methods for analysing different levels of quantitative data have been in co...

International audience; Demographers describe and analyse individual events at multiple levels of observation that range from the individuals themselves to the overall population of interest. In quantitative population studies, one way to streamline investigation is to perform a multilevel statistical analysis using a single model, which improves the accuracy of the estimates and therefore of the results.To that end, this book guides the reader through the first stages of multilevel analysis, from design to implementation, with step-by-step explanations on how to navigate the three most common statistical software environments (Stata®, SAS®, and R). Concrete examples based on census data are provided using an analysis of school enrolment in rural Kenya.Intended for all statistical database users seeking to develop or expand their knowledge of multilevel analysis, this manual details and illustrates the procedures for creating multilevel models and discusses their prerequisites, advantages, and limitations. Suggestions for further reading are also provided.

Models can sometimes fail to run when following particular procedures on specific software environments or following all procedures. To deal with this issue, it is essential to understand the differences between the algorithms applied by the common procedures in SAS, Stata, and R. Thus, this chapter reviews some of the common difficulties encountered with these environments, followed by a look at how to modulate and reuse results. Finally, we will explore some additional avenues for analysis....

Preparing the database is a fundamental step in multilevel modelling and is inarguably the most time consuming due to the often difficult task of finding exogenous explanatory variables. Apart from surveys that collect their own contextual data, no other sources readily provide contextual information covering all the aggregation units both viable for analysis and within the same period. Matching this dual requirement of time and space is sometimes a conundrum, particularly in historical analy...