I would do a multilevel model with children nested within teachers.The only
question left is how to handle the pre-post testing. It could be a rpeated
factor or a growth curve with random intercept and fixed slope. This would
not make nay difference unless there was substantial difference in the time
between pre and post across classes. If so, the growth model would be
preferred. Or you could use the pretest as a covariate. You have plenty of
teachers to estimate the random effects of the teachers. Grade could be used
as a fixed effect at the teacher level if desired. HLM, MLWin, SAS, and now,
I think, SPSS (V12) should handle it. There are other programs but these are
the main ones.

Paul R. Swank, Ph.D. 
Professor, Developmental Pediatrics
Medical School
UT Health Science Center at Houston 


-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On
Behalf Of Harris, Betty A.
Sent: Friday, May 21, 2004 11:36 AM
To: [EMAIL PROTECTED]
Subject: [edstat] Unit of Analysis


Paul S Asked:
How many teachers in each group?

                       Grade
Group          K  1st  2nd  3rd All Grades
# Classes
 Treatment    26  31   21   12     90
 Comparison   19  23   14   15     71
 Total        45  54   35   27    161

# Students tested at both fall and spring
 Treatment   308 407  344  189   1,248
 Comparison  213 237  171  182    803 
 Total       521 644  515  371   2,051

Criterion referenced tests administered to K and 1st graders Norm referenced
achievement tests administered to 2nd and 3rd graders

Betty

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
On
Behalf Of Harris, Betty A.
Sent: Friday, May 21, 2004 8:08 AM
To: [EMAIL PROTECTED]
Subject: [edstat] Unit of Analysis

Hi all, 

We have data from a quasiexperimental nonequivalent control group design. 

A treatment group where the teachers are trained and are implementing a
professional development model that involves group  and individual coaching
to use data to identify their students' component reading skills and target
their teaching interventions to get the missing skills in place. They use
whole group, small group and individual instruction.  

A comparison group of classrooms from demographically similar schools doing
whatever they normally do. 

The students were tested in fall and again near the end of the school year. 

What is the appropriate unit of analysis? And why? 

What if in some of the treatment group students were grouped into
skill-based groups across classrooms for reading instruction?

Thanks, 
Betty 
.
. =================================================================
Instructions for joining and leaving this list, remarks about the problem of
INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to