Advances and Applications in Statistics
Volume 16, Issue 1, Pages 49 - 64
(May 2010)
|
|
EFFECT OF ATTENUATION CORRECTION FOR THE REDUCTION OF BIAS IN PSYCHOLOGICAL MEASUREMENT
Marí¡ Jes?rrera-Fernᮤez, Guillem Feixas-Viaplana and Joan Guತia-Olmos
|
Abstract: This article investigates the impact of bias and measurement error in applied psychological research. In psychology, correlation coefficients based on observational scores may exhibit an effect called attenuation, which is the underestimation of the linear relationship between the variables. In response to this phenomenon, researchers employ the technique of attenuation correction, which, if not applied, may lead to the misinterpretation of data.
This study attempts to show how the simplest version of attenuation correction can provide an adequate approach to error reduction. We worked with two samples - one composed of students and one from the general population - that made up a total of 105 subjects (55.2% in the student group and 44.8% in the community sample). After a preliminary descriptive analysis, both groups were compared using the contrast statistics ?and the Student?s t-test. Using the significant variables, we established a binary logistic regression model to determine the variable association with either group. We subsequently corrected the attenuation by centering and standardizing the variables and again created a binary logistic regression model, which we then compared to the previous model.
The results indicated that the best predictor variables, apart from demographics, are the variables that are obtained from the repertory grid technique. On the other hand, performing an attenuation correction makes the model more complex and slightly increases the total percentage of classified cases. |
Keywords and phrases: attenuation correction, measurement error, standardization, bias, repertory grid technique. |
|
Number of Downloads: 351 | Number of Views: 1402 |
|