Using all of the covariates that are empirically or theoretically thought to be related to special education identification, for each child in the sample, we will create propensity scores representing the likelihood that they would be enrolled in SEP (Class Notes). As in Morgan et al., 2010, we will use multiple matching techniques (i.e., kernel matching, nearest neighbor, and stratification matching).

Once each individual is given a propensity score, we will create “blocks” of individuals in both the treatment (SEP) and control (non-SEP) groups based on their score. The creation of blocks is an iterative process. First, we will need to ensure that there are both treatment and control students in each block. Next, we look at the average propensity scores for each group in every block to make sure they are equal. Then, we compare the average covariate values in both groups in every block to be sure that they are the same. If any of these conditions are not met, we need to create new blocks until they are all met (Class Notes). Once our propensity scores have been created and have met all the necessary conditions, we can move ahead with analysis. For each block, we will estimate the treatment effect. We then calculate a weighted average to obtain an estimate of the overall effect. This is called the weighted average treatment effect (Class Notes). This will be done separately for each outcome of interest.

Don't use plagiarized sources. Get Your Custom Essay on
Just from $13/Page
Order Essay

For an additional comparison, we will create propensity scores and conduct the analysis twice: once with the “control” group being all non-SEP students and the other being students who were classified as special education in kindergarten, but did not attend SEP. This will help us understand if the effects of attending SEP are different when students who received it are compared to students who potentially could’ve received it but were not identified prior to Kindergarten than when they are compared to all non-SEP students. For both sets of analyses within-block comparisons will be run (like in Morgan et al., 2010). This will tell us if the effectiveness of SEP depends on a child’s likelihood of attending.


and taste our undisputed quality.