69
Table 4.2: Results of Р-values of Five Covariates for Acquisitions
|
Estimates |
Exp of Est |
SE of Est |
z Statistic |
p-value | |
|
Growth |
0.335 |
1.398 |
0.228 |
1.468 |
0.140 |
|
Profit |
0.255 |
1.290 |
0.236 |
1.079 |
0.280 |
|
Size |
0.724 |
2.063 |
0.255 |
2.838 |
0.005 |
|
Age |
0.100 |
1.105 |
0.232 |
0.429 |
0.670 |
|
Liability |
-0.327 |
0.721 |
0.260 |
-1.257 |
0.210 |
bankrupt before 1994 are excluded from this research due to the potential bias. All
covariates are redefined as binary variables categorized by the median before model
fitting. Specifically, minimum to median is defined as O and median to maximum is
defined as 1.
4.4 Sensitivity analysis and results
Note that the true value of Kendall’s τ is unknown. In reality, we assume that the
value of τ can be obtained by asking for experts’ opinions. In this research, we applied
different values of τ to conduct a sensitivity analysis. Specifically, we tested Kendall’s
τ = —0.5, 0, 0.2, 0.5, 0.8 with corresponding values of a, as shown in Table 2.2.
To answer the two questions in Section 3.1.2, the results show that some covari-
ates would become non-significant when dependent censoring is taken into account.
For example, “Liability” in the first scenario. Some covariates would become non-