First, we specify a first-order autoregressive error process and test the null hypothesis that the errors are not first-order autocorrelated. |
|
The residuals are neither autocorrelated nor heteroscedastic and display no significant deviations from the normal distribution. |
|
However, Lehavy and Sloan report that changes in investor recognition are also autocorrelated and that after controlling for this autocorrelation, a negative relation emerges. |
|
Diagnostic tests for serial correlation determined that the residuals from the regression model were autocorrelated and therefore an ARIMA analysis was used. |
|