The impact of the banking sector development on the financial performance of the communication sector in sierra leone

4.10 Diagnostic Testing 
ARDL model provide objective results on a smaller set of time series related 
data and for the result to be robust, certain diagnostics test needs to be 
conducted to determine the validity of the results and also to ensure the results 
are statistically robust. However, the test for stability, heteroscedasticity, 
misspecification, serial correlation and residual normality are needed to be 
conducted to ascertain if the model is free from biases and can provide a 
satisfactory result. If the tests produce satisfactory results, we are in good 
position to use it for analysis. 
4.10.1 Test for Stability 
ARDL model is highly sensitive to structural break and also the use of financial 
related time series that is sensitive to events globally, the need to analyse the 
stability of the coefficients is of major significance in order to assess the 
stability of the long-run and short-run coefficients. The test for stability 
proposed by (Brown, Durbin, & Evans, 1975) are the CUSUM and CUSUMSQ 
test and can be conducted for stability. 
The CUSUM test is based on the cumulative sum of the recursive residuals 
and the CUSUMSQ test is on the cumulative sum of squared recursive 
residuals (CUSUMSQ) and they are of graphical nature whereas the residuals 
are updated recursively and plotted against the break points for the 5% 
significance line. The concept of the CUSUM test provides, where the 
cumulative sum of recursive residuals is plotted against the upper and lower 
95 % confidence bounds and the same goes for the CUSUMSQ. If results of 
the test show a range within 5% significant level, indicating that the long-run 
and short-run coe
fficients are stable. However, both tests analyses if the 
residuals do not significantly diverge from its mean value by imposing parallel 
critical lines on a significant level of 5%. This is illustrated on figure 2 below.