Z
P
ERM
0.0568
0.045
1.26
0.208
0.3315
0.262
1.26
0.207
ddLOGSI
ZE
-0.7491
0.160
-4.67
0.000
0.1863
0.340
0.55
0.584
ddROA
-0.2463
0.337
-0.73
0.466
-0.6565
0.543
-1.21
0.227
dGD
0.2738
0.195
1.40
0.161
-0.1779
0.256
-0.69
0.488
ddLEV
-0.8647
0.412
-2.10
0.036
0.0506
0.811
0.06
0.950
PS
0.0013
0.000
2.81
0.005
-0.0010
0.003
-0.34
0.732
GS
-0.0628
0.124
-0.51
0.613
-0.7094
0.424
-1.67
0.095
Constant
-0.0996
0.034
-2.89
0.004
1.9636
0.311
6.30
0.000
Number of Observation: 165
Number of Groups: 33 P = 0.0000
R
2
(within) = 0.19 Wald
x
2
=37.15
Number of Observation: 165
Number of Groups: 33 P = 0.0000
R
2
(within) = 0.10 Wald x
2
= 11.25
Note: The term “
d
”
indicates the difference between the relevant variable.
The validity of
t
and
F
statistics, R
2
and confidence intervals
are affected if there is heteroscedasticity, autocorrelation and
correlation between units. Therefore, if the model has at least one of
variance, autocorrelation and correlation between units, resistant
predictors should be used (Yerdelen Tatoğlu, 2013: 24
2). Since the
critical values were exceeded in the tests for the heteroscedasticity,
autocorrelation and correlation between units assumptions regarding
the models used in the study, standard error-resisting prediction
models were used which gave more consistent results considering
these assumptions.
It is common to rely on durable standard errors to provide
valid statistical inferences when the assumptions of the basic
regression model are violated. The most common of the alternative