Improved estimations of Standard Errors

If individual values are similar within the areas (in statistic language: residuals are dependent) the number of “statistical individuals” (in statistic language: sample size) becomes reduced.

Think! Two identical twins are only one “statistical individual”….Isn’t it?

In ordinary regression techniques two identical twins are two individuals, but in multilevel regression only one. Ordinary regression techniques overestimate the sample size giving smaller standard errors (the higher the samples size the lower the standard error).

Exploring variance

How much of the individual differences are between the areas (how much of the teeth health variance is because of the mouths?)….1%?... 10%...50%?. This measuring of clustering is obtained by the intra-class correlation.

To know if individual associations vary between the areas (in statistic language: random slopes). The slope in a regression analysis is a measure of association that could be different in different areas.