How do I analyze *groups* of students' academic performance over different school start times?

[LLP1]

Hi,

I'm helping a friend's daughter on a high school research project that aims to examine academic performance (i.e., mean test scores and overall class % attendance rate) across different school start times, adjusting for socioeconomic status (SES).  The data is grouped by grade level and school, so I do not have individual level data.  I do, however, have a breakdown of scores & % attendance broken down by SES (disadvantaged vs. not disadvantaged), as well as the number of students in each group/subgroup.

My dependent (outcome) variables that I wish to compare are: 

--mean reading test scores for a grade level & school  (and # of students in that grade level & school)

--mean reading test scores for a grade level, school and SES  (and # of students in that grade level, school and SES stratum)

--mean math test scores for a grade level & school  (and # of students in that grade level & school)

--mean math test scores for a grade level, school and SES  (and # of students in that grade level, school and SES stratum)

--% attendance rate for a grade level & school (and # of students in that grade level & school)

--% attendance rate for a grade level, school, and SES (and # of students in that grade level, school and SES stratum)

I cannot adjust by race/ethnicity or gender, since I don't have individual level data.

Would this be analyzed using ANOVA or a linear (or Poisson for the attendance rate?) regression analysis?  Or neither of these?

If ANOVA or regression, how would I write the SAS code to take into account the grouping of data and number of students in each group?

Thanks for any help you can provide,

--Laura

[PaigeMiller]

The data is grouped by grade level and school, so I do not have individual level data. 

You can't do any type of statistical analysis unless you have N in each group, the MEAN in each group, and the standard deviation in each group. And by group, I mean (in one case) grade level & school (you need the N, MEAN and standard deviation for each grade level & school group), because that's the comparison you are doing, and in other cases, the group is grade level, school and SES (you need the N, MEAN and standard deviation for each grade level, school and SES group), because that's the analysis you are doing. Do you have this?

[LLP1]

I have the mean score and the N, but not the standard deviation for each of my groups/subgroups.  I will check on the public website to see if they provide standard deviations, though I'm not optimistic.  (sigh)

Thanks,

Laura

[LLP1]

I’ve submitted a non-confidential data request for the standard deviations of each of my groups/subgroups and am waiting to hear back. These requests are reviewed the 1st Monday if each month, so I’m thinking sometime next week. Thank you for your help, and I’ll be back!

[PaigeMiller]

I'm sorry, but after my original message, I decided that even group standard deviations would not help. I said:

by group, I mean (in one case) grade level & school (you need the N, MEAN and standard deviation for each grade level & school group), because that's the comparison you are doing, and in other cases, the group is grade level, school and SES (you need the N, MEAN and standard deviation for each grade level, school and SES group), because that's the analysis you are doing

This would not result in a true ANOVA. You would really need the individual data to do a true ANOVA. Having group variances might let you do SOME of the t-tests, but unless you had standard deviations for all t-tests you want to do, you still couldn't do them all and even if you did them all, that's not a true ANOVA.

[LLP1]

Oh, that’s too bad. It did seem a bit too good to be true that we could do this without the individual student values. We’ll have to settle for a “descriptive analysis” that simply compares the mean test scores across each grade level and SES by school start time. If it looks like there’s a difference by school start time, perhaps she can make the case to gain access to and analyze the individual level data if someone within the school system will sponsor her research?
Regardless, I really appreciate having someone with such expertise readily available to consult on this! Thank you for your help!