1

New to LME's and wondering if someone might be able to help me with this question. Does the application of a random block in an LME apply to both the dependent variable and the independent variable(s)? For instance if my model is:

mod <- lme(concentration ~ growth rate, random = ~1|site, data = all.data)

does this application of a random block assume an intercept that is different for each site, telling my model it will have various responses dependent upon variable baseline concentration levels per site, or baseline concentrations and individual growth rates?

Stefan
  • 6,431
nordic5
  • 11

1 Answers1

0

With your syntax, the intercepts vary across sites. Your syntax will give you a model of the form:

$$concentration_{is} = B_{0s} + B_{1} rate_{s} + \epsilon_i$$

where $s$ subscript indicates the site (level-2) index and the $i$ subscript indicates the observation (level-1) index. The term $B_{0s}$ is the intercept that varies across sites $s$. In this model the level-2 residuals (i.e., the varying intercepts) are assumed to be uncorrelated with the level-1 residuals (i.e., $\epsilon_i$, the errors in prediction at the observation level). Note that not knowing about your context I'm assuming growth rate is a predictor that varies at the site level (hence the absence of an $i$ subscript), but it's the same if the predictor varies at level 1.

Patrick Coulombe
  • 2,810