Sunday, October 14, 2007

Thesis Update

Have completed all the major components of the analysis required, and am now in the process of writing up.

The analysis components included performing a Quandt Test, a Chow Test, a Switching Regression, and a cluster analysis, using both fixed and industry effects to account for the different industries and their effects upon the overall function which this thesis project uses. The Chow test was selected in order to determine breakpoints, which varied from the original thesis scope in that the original breakpoints for the cycles to perform the analysis were set arbitrarily in order to evenly distribute the data across the cycles. Furthermore, the clustering analysis, when performed upon 19090-1995, 1996-1999, and 2000-2006, reveals that there are originally 3 clusters, which diverges into 5 during 1996-1999, and then settles down to 4 clusters. The clustering analysis, done via setting the fuzziness exponenet m to 2, using the partition coefficient method 1/c, therefore indicates that there is a divergence, and then convergence, across the industries during the cycles. The divergence between 1990-1995 and 1996-1999 can be explained by the fact that the middle cycle, which was determined by a Chow test, can be called a transitional period (such as that in Dr Poon's paper), and also reveals intra- and inter- differences between the clusters.

The other major component, that of switching regression, indicates that the model used is an endogenous model, and not exogenous, as the rhos for both cycles of the switching regression significantly differ from 0, which indicates the aforementioned (Bertschek et al, 2005).

No comments: