Psychiatric Bulletin (2006) 30: 354. doi: 10.1192/pb.30.9.354-b
© 2006 The Royal College of Psychiatrists
Psychiatric Bulletin (2006) 30: 354
© 2006 The Royal College of Psychiatrists
Statistical assessment of MMSE scores
Alastair Willis, Medical Officer (Mental Health)
W
hi Oranga, Nelson Hospital, Private Bag 18, Whakat
7001,
New Zealand, e-mail:
Alastair.Willis{at}nmhs.govt.nz
It is disappointing that the interesting study by OLoughlin &
Darley (Psychiatric Bulletin, April 2006, 30, 131134)
was let down by the use of inappropriate statistics. Since scores on the
Mini-Mental State Examination (MMSE) constitute data that are ordinal in
nature, it is not appropriate for the mean to be presented as a measure of
central tendency. For the same reason, it is not appropriate for standard
deviation to be offered as a measure of dispersion. Use of the median and
interquartile range (IQR) would have been more appropriate. Similarly, use of
the t-test as a test for difference between the two groups was ill
considered because MMSE scores in both study populations are negatively
skewed. The authors should have used a non-parametric test for difference such
as the MannWhitney U-test.
For the record, the median MMSE score was 20 (IQR 1624) in the 1996
sample and 22 (IQR 1925) in the 2003 sample. Running the authors
data through a MannWhitney test on Stat Crunch (available at
http://www.statcrunch.com)
still finds a significant difference between the two groups
(P=0.0037).