Description

Many of the relationships of interest in the behavioral and social sciences are not necessarily linear in nature, and it has been increasingly recognized over time that a linear specification (e.g., linear regression) may be leaving important aspects of these relationships unexplored. One popular analytical technique for the modeling and analysis of these specifications is polynomial regression. However, the application of polynomial regression when the constructs of interest are not perfectly measured is not without perils. In this research, we examine the performance of a variety of analytical techniques that do take measurement error into account, and compare their accuracy with regards to parameter estimation. Preliminary results indicate the use of OLS (Ordinary Least Squares) should be abandoned in favor of the alternative techniques examined here.

Share

COinS
 
Aug 10th, 12:00 AM

Polynomial Regression and Measurement Error: Implications for IS Research

Many of the relationships of interest in the behavioral and social sciences are not necessarily linear in nature, and it has been increasingly recognized over time that a linear specification (e.g., linear regression) may be leaving important aspects of these relationships unexplored. One popular analytical technique for the modeling and analysis of these specifications is polynomial regression. However, the application of polynomial regression when the constructs of interest are not perfectly measured is not without perils. In this research, we examine the performance of a variety of analytical techniques that do take measurement error into account, and compare their accuracy with regards to parameter estimation. Preliminary results indicate the use of OLS (Ordinary Least Squares) should be abandoned in favor of the alternative techniques examined here.