Covariance Analysis of Heterogenous Group Data

Mohammad Masjkur, M. Sjarkani Musa, . Aunuddin, Oetit Koswara

Abstract


This paper discussed the important of concomitant variables in regression analysis. Particularly, it was shown that without considering concomitant variables none of the combined regression models was reliable. In fact, the highest coefficient of determination R' was obtained from Mediterran soil group of data (65.8%). although R' for each specimen varied from 72.8% to 97.7%. Considering K-soil contents as concomitant variables, it was shown that the coefficient of determination increased by 18 2% to 85.3%. The resulting R* for each model was ranging from 67.1% to 93.6%. Another important finding from the research was the incorporation of concomitant variables resulted in more symmetric data distributions, smaller variances, and substantial increase in the regression slopes.

Full Text:

PDF