|
|
Random Groups Method
|
Another
point of statistical contention is the "random groups method." Weller and Weller, the researchers who have
done the most recent menstrual synchrony studies, construct a "random
control group" in the following manner:
"...using a table of random
numbers, the participants in the experimental group are randomly paired with
each other. The synchrony scores
obtained from the real pairs are then compared with those from the
statistically constructed random control group, by means of paired t tests."
Weller and Weller go on to state that:
"Remarkably, in every study,
including McClintock's original study and a later study of estrous synchrony in
rats, when such comparisons were made the experimental group was always
significantly more synchronous than the random control group."
Schank replies as follows:
"The [random control group] procedure
is invalid because the actual sample and constructed sample are not
independent. For example, if all
subjects are randomly repaired, the mean cycle length and variance will be
identical for the two groups."
Schank then goes to on critique a specific
example, "Weller and Weller (1993) [applying] the random groups method to
their dormitory samples":
"For the real data, they report
N=108 pairs, but their "randomly constructed" control group, the
constructed N=93 pairs. Why the
asymmetric sample size? The mean
synchrony score of real data was approximately 7.0 days and for the randomly
constructed control group, 7.3 days...They reported p < 0.01 using the
Mann-Whitney U-test. However, 7.0 days
was the expected onset difference.
Thus, they found a significant difference between two non-synchronous
samples! The random group method, far
from being "The most serious criticism of Schank's contentions..."
further illustrates the unreflective misuse of statistical methods in their research."
Click here for the next
MOLTXPERIMENT.
X Who would you invite to a random control group?
X
Menstrual Synchrony Index
X Return to MoltXibits
X Return to Main Page
X Contact MOLT