|
Activity Number:
|
50
- Methodological Developments and Implications for Social Scientists
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Social Statistics Section
|
|
Abstract #313392
|
|
|
Title:
|
A Reparametrized Linear Model That Parses Qualitative and Quantitative Characteristics of a Set of Predictors Relative to a Criterion
|
|
Author(s):
|
Ernest Davenport* and Mark L Davison
|
|
Companies:
|
University of Minnesota-College of Education & Human Development and University of Minnesota
|
|
Keywords:
|
Quantitative;
Qualitative;
Linear Models;
Math Courses;
Math Achievement
|
|
Abstract:
|
Davison & Davenport (2002) reparametrize any regression equation into Level & Pattern. Level is the mean & Pattern is the covariance of a participant’s predictor scores to regression weights predicting higher scores on the criterion. This study illustrates how these 2 variables are related to quantitative & qualitative aspects of the predictors relative to the criterion. Level represents quantitative predictability, while Pattern is useful to the extent the predictors relate to the criterion in qualitatively different ways. This allows one to conduct quantitative & qualitative analyses simultaneously. Our illustration uses math course-taking relative to math achievement. Data are from NCES’s High School Longitudinal Study of 2009 which contains math test scores and transcript data from a nationally representative cohort of ninth graders in 2009. Thirteen variables based on the 67 SCED math courses were generated based on highest math course taken (X3THIMATH). If quantitative information in the courses is sufficient, Level is a sufficient predictor. If the courses are differently predictive, qualitative information in the predictors should also be considered.
|
Authors who are presenting talks have a * after their name.
