Topic: Title: A comparison of the power of tests for heteroscedasticity in non-linear econometrics models Authors:  Iyabode Favour Oyenuga - The Polytechnic Ibadan (Nigeria) [presenting]
Benjamin A Oyejola - University of Ilorin (Nigeria)
Abstract: There are several methods of estimating the power of test for heteroscedasticity in a single equation econometrics model. We investigate the power of White and Breusch-Pagan test for heteroscedasticity using three non-linear models: Cobb-Douglas, constant elasticity of substitution and exponential mode. Contaminated heteroscedasticity samples were generated using uniform distribution to generate data for capital $K$ and labour $L$ which were used to generate the output $Y$.The parameter estimates for the three models were obtained. The new set of parameters were used to compute the residuals which were used as the dependent variable for the auxiliary regression. The sample sizes used for the simulation were 10, 30, 50, 100, 150 and 200 with 10,000 replications. The results show that at 5$\%$ level of significance for the White test, as the sample sizes increase the power of the test decreases for the three models. In Breusch-Pagan test, as the sample sizes increase the power of the test increases for exponential model but its power is weak for the remaining two models.