MENENTUKAN KOEFISIEN REGRESI EKSPONENSIAL DENGAN METODE KUADRAT TERKECIL SEDERHANA DAN METODE KUADRAT TERKECIL BERBOBOT

Abstract

Relation between two variables x and y are not always linear, but also non-linear. Scatter diagram from non-linear relationship will show a pattern of data points that can be approximated by an exponential curve. The process of determining an exponential curve that best fits a data set called an exponential regression. In order to get the best an exponential regression curve was used a standard least square method with logarithm transformation. However, using this logarithm causes error change from error in the variable y to the error in the variable ln y. To overcome this problem, Glaister Internat. J. Math. Ed. Sci. Tech.. 38 (2007): 422-427 proposes a least square method that is based on applying a weighting in the standard least square method that is based on the error in the logarithm of a variable. This method is called a weighted least square method that we discuss in this article. By comparing these two methods in the case of an exponential data, the performance of the weighted least square method is better than that of the standard least square method.