Kaiser, H. F.

(1974). An
index of
factorial
simplicity.
Psychome-
trika, 39,
31–36.
Kaiser, H. F. (1974). An index of factorial
6043 citations
in Google
Scholar as of
simplicity. Psychometrika, 39, 31–36.
4/1/2016

6043 citations in Google Scholar as of 4/1/2016

Psychometric Society, Asheville, NC, July, 2016

 The Miscitation Story Begins

Kaiser, H. F.
(1974). An
index of
factorial
simplicity.
Psychome-
trika, 39,
For all articles that have appeared in Psychometrika since its
31–36.
first volume in 1936, this highly cited paper by Henry Kaiser is
6043 citations
in Google second only to Cronbach’s massively cited 1951 article on
Scholar as of
4/1/2016 coefficient alpha.
This Kaiser article suggests a normalized quartimax criterion,
bounded between zero and one, to index the simplicity of the
factor pattern for a given factor analysis.
Unfortunately, almost all but a very small handful of the
citations to Kaiser (1974) appear to be in error.
 The Kaiser-Meyer-Olkin Measure of Sampling
Adequacy: The Root of the Problem
Kaiser, H. F.
(1974). An
index of In his Psychometric Society Presidential Address in 1970 (“A
factorial
simplicity. second generation Little Jiffy,” Psychometrika, 1970, 35
Psychome-
trika, 39,
401–415), Kaiser added a few embellishments to his
31–36.
well-known approach to factor analysis that Chet Harris had
6043 citations
in Google disdainfully labeled “Little Jiffy”:
Scholar as of
4/1/2016 “principal components with associated eigenvalues greater than
one followed by normal varimax rotation.”
One such addition in this 1970 paper was a “measure of
sampling adequacy” (MSA) that was intended to reflect
whether it was reasonable to proceed with a factor analysis in
the first place.
Kaiser attributed this MSA to work he was doing at the time
with Professors Meyer at Loyola (Chicago) and Olkin at
Stanford.
Kaiser, H. F.
(1974). An
index of This MSA is now commonly referred to as the
factorial
simplicity. Kaiser-Meyer-Olkin Measure of Sampling Adequacy
Psychome-
trika, 39, (KMO-MSA).
31–36.

6043 citations It is calculated routinely, for example, in the heavily-used SPSS

in Google
Scholar as of and SAS factor analysis programs; also, functions for
4/1/2016
KMO-MSA appear in R, such as in the psych package
developed by Bill Revelle.
The particular KMO-MSA computed by SPSS and various R
functions, is not the exact same measure given in Kaiser
(1970); instead, it is a modification meant to improve stability
given in a 1974 Kaiser and Rice paper in Educational and
Psychological Measurement (“Little Jiffy, Mark IV,” 34,
111-117).
 The Citation Mixup Arises

Kaiser, H. F.
(1974). An Over the last four or so decades, when the value of the
index of
factorial KMO-MSA index is reported (usually taken directly from SPSS
simplicity.
Psychome- output), the reference given for it is most often Kaiser (1974),
trika, 39,
31–36.
the factorial simplicity paper, and not the correct citation of
6043 citations Kaiser and Rice (1974), the Version IV Little Jiffy article.
in Google
Scholar as of One reason for Kaiser (1974) becoming the inappropriately
4/1/2016
highly-cited paper it has developed into, may be due in part to
an article by Charles Dziuban and Edwin Shirkey from
Psychological Bulletin, also in 1974: “When is a correlation
matrix appropriate for factor analysis” (81, 358–361).
the following sentence appears on page 359: “Kaiser’s (1974)
present calibration of the [KMO-MSA] index is as follows:”
No reference, however, appears for Kaiser (1974) in the
bibliography for Dziuban and Shirkey (1974) but one is given
for an “in press” piece by Kaiser and Rice.
Kaiser, H. F.
(1974). An
The mixup of using Kaiser (1974) for Kaiser and Rice (1974)
index of
factorial
may be a miscitation phenomenon that is difficult to correct.
simplicity.
Psychome- Two popular SPSS-related user manuals, for example, make
trika, 39,
31–36.
this citation error: the SPSS Survival Manual by Julie Pallant,
6043 citations and Discovering Statistics Using SPSS by Andy Field; the same
in Google
Scholar as of citation error is also made in other books by Andy Field, such
4/1/2016
as in Discovering Statistics Using SAS.
In hindsight, it is surprising that such a miscitation wasn’t
caught earlier by an author of these secondary SPSS user
manuals.
Not a single statistical package (SPSS, SAS, SYSTAT, Matlab,
or R) computes Kaiser’s index of factorial simplicity,
irrespective of how good or bad the index might be.
Several statistical packages do, however, compute KMO-MSA
(SPSS, SAS, and Bill Revelle’s psych package for R).
 Comments by Bill Revelle

Kaiser, H. F.
(1974). An
index of
factorial
simplicity. When Bill Revelle was asked why he didn’t include a function
Psychome-
trika, 39, for Kaiser’s index of factorial simplicity in his psych R package,
31–36.

6043 citations
he said that he instead included a function for what he thought
in Google
Scholar as of
was a better index (of factorial complexity) due to Richard
4/1/2016 Hofmann (“Complexity and simplicity as objective indices
descriptive of factor solutions,” Multivariate Behavioral
Research, 13, 1978, 247–250).
He also noted that he was never tempted to use Kaiser (1974)
in reference to his KMO-MSA function; he gave the three
(correct) citations for it of Kaiser (1970), Kaiser and Rice
(1974), and Dziuhan and Shirkey (1974).