The PAMS (Profile Analysis via Multivariate Statistics) Perspective on Educational Survey Data via Correspondence Analysis

• In Event: Profile Analysis via Multivariate Statistics for Continuous and Categorical Data

Abstract

Objective: The current study uses the profile pattern approach to interpret dimensions in correspondence analysis for the purpose of finding a new perspective on complex educational data.

Theoretical framework: Correspondence analysis allows us to identify common principal axes (dimensions) for rows and columns and to estimate dimension coordinates by projecting rows and columns onto those same principal axes. Because rows and columns contain different information, however, the interpretation is different. By applying the PAMS approach to this information, we analyze the dimension coordinates as profile pattern information. This allows us to find common themes, in this case, among different stakeholders’ view of their schools.

Method: 150 questions from three survey versions (parents, teachers, students) are categorized into how well the survey taker believe their schools fared in four key areas: Academic Expectations, Communication, Engagement, and Safety & Respect. A twenty-three (parents, teachers, students x school types) by sixteen (4 key areas x 4 scale points) contingency table is analyzed.by CA to identify row/column dimensions. Then, dimension coordinates are converted into correlation coefficients to enhance interpreting row/column profile in terms of dimension profile patterns.

Data sources: Survey results from parents, teachers, and students who completed the 2010 New York City School Survey (N=908,407).

Results: Since a 23 by 16 contingency table was analyzed, the possible number of dimensions must be 15 because of the rule min(23,16)-1. However, only two dimensions were included for interpretation because the first (89%) and the second (7.3%) dimensions accounted for the total variation more than the average 6.7% (=100/15). The correlations between row profiles and the first two row dimensions were computed. The first row dimension was labeled as Communication and Safety/Respect Profile and the second row dimension as Academic Expectation and Engagement Profile. The row profiles for Parents ,irrespective of school types (Elementary, Middle, and High), were highly but positively related with the first row dimension profile (.72≤r≤1.00), whereas most of Students irrespective of school types, except YABC schools, were negatively related with the first row dimension profiles (-.99≤r≤-.38). These results imply that parents are concerned more about communication and safety/respect issues than students. The rest of the results will be included in the full paper.

Scientific/Scholarly significances: At the time of the takeover of the NYC schools by the mayor in 2002, one of the biggest concerns was that the lack of decentralized authority would prevent parents, students, and teachers at local schools from being heard. The yearly school survey is an attempt to let these voices be expressed. However, after collecting this data, the types of analyses that have been run by the city have been simplistic at best. One of the many areas that has gone unexamined is how the three groups’ (parents, students, and teachers) evaluations compare on the key ingredients that make good schools. Applying a profile pattern interpretation to this data allows us to recognize and explore patterns of expectations and frustration that can facilitate discussions between these three groups and the city that runs the schools.

Authors

• Jennifer Lord-Bessen, Fordham University

• Se-Kang Kim, Fordham University