Page 22 - Contributed Paper Session (CPS) - Volume 8

CPS2151 Sarah B. Balagbis
(80) time points are considered resulting to a panel with 5,600 data points. The
first independent variable (X1) is generated from a uniform distribution from
the interval (20, 30). The second independent variable (X2) is generated from
N(10,1).
The dependent variable (Y) is generated by first simulating the
independent variables and the error terms (at). Eighty (80) at’s are simulated
from a standard normal distribution. Given at’s and setting ρ=0.5, the error
component = (−1) + is computed. Note that we need the initial

value . We set = 0 which is tantamount to setting the first value 11 =
0
0
. All the other value of for t=1 will be initialized by the previous value in
1

the data layout. Using the simulated independent variables X1 and X2, and the
computed error component , and setting = 1.2, =0.4, and =0.6, the
2

0
1
dependent variable = + + + is computed. This
2 2

0
1 1

constitutes the dependent variable without perturbation.
Panel heterogeneity is introduced into the system by replacing the Y values
of the randomly chosen panels for all T =1, 2, …, 80 by setting = 0.8, =1.4,
0
1
=2.1. Structural change is introduced into the system by choosing time
2
points at the start, middle, end, and time points located in all 3 locations (start,
middle, and end) and changing the Y values for these time points for all N =
1, 2, …, 70 by setting ρ=0.8. Panel heterogeneity and structural change are
incorporated by combining the two strategies above and by setting = 0.8,
0
=1.4, and =2.1, and ρ=0.8.
2
1
Twenty data sets without perturbation are simulated in the study and for
each of these, 5% (≈280 observations) and 10% (≈560 observations) panel
heterogeneity or structural change are incorporated. For the structural change
and a mixture of structural change and panel heterogeneity, four scenarios are
considered: (i) perturbations at the start; (ii) perturbations at the middle; (iii)
perturbations at the end; and (iv) perturbations at the start, middle, and end,
i.e., spread all throughout the 3 locations (Table 1).
For the 5% panel heterogeneity, 3 panels are randomly selected while for
10% panel heterogeneity, 7 panels are randomly selected. Four time points are
selected at each different location (start, middle, end, throughout the 3
locations) for the 5% structural change, and 8 time points are selected for the
10% structural change. For a mixture of panel heterogeneity and structural
change, 14 panels are randomly chosen and 20 time points at each location
are selected for the 5%, while for the 10% panel heterogeneity and structural
change, 28 panels are randomly chosen and 20 time points at each location
are selected. A total of 360 data sets with perturbation will be used in the
study:

11 | I S I W S C 2 0 1 9