Meeting 2010 February 18

Summary

Discussed questions that Ellen and Peter had about the analysis plan.

• There are only two outcomes, m_tgmean_36_80m and m_ttp_trust_80m. Both of these are teacher ratings about the principals.
• Gap scores may need to be modeled using their actual continuous value rather than using a dichotomous variable based on a cut off. They can predict teacher ratings, but should not be used to predict the principal's rating of himself. We may use the non "80m" variables to calculate the gap variable, since the mean variables not using the 80% cut off were the values that were given in the feedback reports.
• We may need to think about a gap between an individual principal's ratings of himself and the district (teacher rating?) average, as well as the gap between the principal's self rating and the teachers' rating of him, which is discussed above.
• We will give standardized effect sizes at time = 12 months.
• We will give the equation of the model we are using in our analysis report (or in the analysis plan) to clarify for the investigators what model we are actually using.
• We may want to account for variance of the teacher ratings in a given school. Ben will think about this. Len wants to look at this variance as an outcome.
• Len, Ellen, et al plan to meet to choose important covariates. They plan to let us know in one or two weeks(?).
• AERA abstracts are due April 1.

Meeting 2010 February 8

• Discussed questions that the rest of the group had regarding the analysis plan.
• Established that in the second wave, feed back reports were distributed at the end of August, and the Wave 2 data collection started in September.
• To think about whether eventually we should take into account in our models whether the amount of time the principal has the feedback report before he is next evaluated, CEPI will create a variable on every wave (missing at wave 1) for every principal giving this amount of time. JoAnn will do summary stats to see how much this variable varies for different principals.
• JoAnn will do some comparisons between participant and non-participants (those who only have baseline data) to see if they differ systematically.
• Percent of students on free lunch program given as the most important covariate.
• We will try to schedule another meeting next week.