Yes, the short answer is YES, BUT there is always the possibility of extraneous factors being introduced when dichaotomies are factored (historically, these have been called "difficulty factors"). The short and sweet of it is that with dichotomies it is likely that several variables may have similar marginal splits (or item difficulty, or proportions of endorsement, etc.). as Gorsuch showed in his 1974 _Factor Analysis_ (p.250), items with similar splits will tend to correlate with each other regardless of content. So, when you run the factor analysis, check the means of the items in each factor, and check the content as well, to make sure that the factor makes sense. If the means of the items comprising a factor are all very similar, it may be a spurious, or difficulty, factor.

Gorsuch, R. L. (1974). _Factor analysis_. Philadelphia: W.B. Saunders.

HTH

Max Martin ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^ SAN ANTONIO SPURS RULE! ^ 1998-99 NBA CHAMPS ^ Go! Spurs Go! ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^ Dr. Max R. Martin ^ Senior Evaluator ^ Department of Research, Evaluation ^ Info Systems & Technology ^ Edgewood ISD ^ 4127 Eldridge ^ San Antonio, TX 78237 ^ Vox: 210-431-9120 FAX: 210-433-5134 ^ e-mail: mmartin@edgewood-sa.k12.tx.us

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

----- Original Message ----- From: Hector E. Maletta <hmaletta@OVERNET.COM.AR> Newsgroups: bit.listserv.spssx-l To: <SPSSX-L@LISTSERV.UGA.EDU> Sent: Friday, November 19, 1999 10:29 AM Subject: Re: Factor Analysis question

> This has come up a number of times. The short answer is YES. Dichotomous > variables can be treated as interval variables, and factor analysis can > be used for analyzing them. > On the other hand, factor analysis will not search for subscales based > on content, but based on the observed correlations between items. High > loadings in one factor may be a feature of questions related to more > than one subject. > > Another approach would be to consider blocs of questions separately, > taking those questions that you a priori consider as related to research > design, say, or statistics, and analyzing the reliability of the > component items. SPSS Reliability analysis can do that. > > Hector Maletta > Universidad del Salvador > Buenos Aires, Argentina > > "Raymond M. Zurawski" wrote: > > > > We have developed a test of knowledge of statistics and > > research design that is comprised of 55 multiple choice items. > > We would like to be able to determine whether there are > > subscales on this measure (e.g., knowledge of statistics, of > > research design, etc.). > > > > Is there a way to perform a factor analysis on the individual > > items given that these items are scored dichotomously (i.e., > > correct-incorrect)? > > > > Thanks in advance for any suggestions. > > > > Raymond M. Zurawski, Ph.D. > > Associate Professor of Psychology > > St. Norbert College > > 100 Grant Street > > De Pere, WI 54115-2099 > > Voice: 920-403-3202 > > Fax: 920-403-4033 > > Email: zurarm@mail.snc.edu > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^ SAN ANTONIO SPURS RULE! ^ 1998-99 NBA CHAMPS ^ Go! Spurs Go! ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^ Dr. Max R. Martin ^ Senior Evaluator ^ Department of Research, Evaluation ^ Info Systems & Technology ^ Edgewood ISD ^ 4127 Eldridge ^ San Antonio, TX 78237 ^ Vox: 210-431-9120 FAX: 210-433-5134 ^ e-mail: mmartin@edgewood-sa.k12.tx.us ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^