The following datasets are offered
as a means for teaching some of the advantages
of Multilevel and Hierarchical Linear Modeling.
Each of the datasets are offered in either S-PLUS
format or in a format suitable for import into
HLM (c.f., SPSS level structure design). Feel
free to use these datasets at your leisure. The
full citation for referencing these datasets would
be:
Roberts, J. K. (2005, April).
Datasets illustrating speicifc strengths of
hierarchical linear modeling. Paper presented
at the annual meeting of the American Educational
Research Association, Montreal. |
- All of the datasets are available below by clicking on the appropriate link. The datasets are zipped
together and will need to be unzipped to be used.
|
- Dataset 1 - Strength of HLM over OLS. In this
dataset, users are shown that sometimes the result
of an HLM analysis can be exactly the opposite from
the results of Ordinary Least Squares (OLS).
|
- Dataset 2 - Fixed versus Random Coefficients and
Level-2 Predictors. The design of this dataset is
to show how adding a random coefficient to a model
dramatically increases data fit even though the
coefficients to the slope coefficient are not statistically
significant. A level-2 predictor is also added to
show the mediation of this effect.
|
- Dataset 3 - Repeated Measures Trends - In this
dataset, the linear trend in the HLM repeated measures
model is illustrated through 10 measurement occasions
nested inside 10 individuals. This dataset illustrates
the linear treand and shows how different rates
of growth can be mediated by second-level covariates
added to the model. Coming Soon.
|
- Dataset 4 - 3-Level Models - This dataset illustrates
how neglecting to consider an additional level in
a model could potentially lead to a misinterpretation
of the data. In this case, the addition of a third
level shows that responses at the classroom level
(level-2) are actually a by-product of the school
from which they are drawn. Coming Soon
|
