Abstract
Srinivasan (1970) proposed a two-sided Kolmogorov-Smirnov type test
for testing the goodness of fit of exponential and normal distributions
with parameters unspecified. For the exponential distribution, this
test was reported to have surprisingly better power than a similar
test proposed by Lilliefors (1969). We show in this paper that the
critical values given by Srinivasan (1970) are too low and the power
calculations are incorrect. We give the correct critical values for
$n = 3(1)15(5)20$ and the correct powers for the exponential null
distribution for $n = 5, 10$ and 20 and for the test sizes given
by Srinivasan (1970).
Users
Please
log in to take part in the discussion (add own reviews or comments).