Gian wrote <<< thank you for you patient! >>> You're welcome. <<< Just before going on, I want to make a precisation about the questionnaire I administered: it asked to respondents to "rate each item depending on how well it described their project leader on a 1 to 9 scale".
Now I would like to know: 1) along which dimensions respondents describe their leaders (the relationship between variables). >>> Factor analysis is good for figuring out relationships among variables. But this may not be how respondents describe their leaders. FA asks "Are there latent variables that account for much of the variance in the questions?" But I am not sure this is what you want.....you may want to use multidimensional scaling, to do this, you would first convert the data to similarities (or dissimilarities) then, MDS would try to figure out what dimensions people are using to group people. You may, on the other hand, want cluster analysis, which asks whether there are groups of leaders who are similar to each other. <<< 2) how much the leaders they described score on the average in each dimension. In addition (and this is my main problem, by now), I would like to calculate how much the leaders they described score e.g. on the "competence" dimension. Factor analysis told me that the three items listed above correlate nicely and thus form a factor. Now, how much the leaders have been described as "competent", by the respondents? >>> Factors always (AFAIK) have means of 0 and standard deviations of 1. And, if you use an orthogonal rotation (such as varimax) the factors cannot be correlated. Eventually, I would like to sort factors dependently on the mean score they got -- I would be able to say something like "the respondent's leaders have been described along these dimensions (e.g. - "leader's competence", "leader's friendliness" [...]) and among these dimensions items correlated to "leader's competence" got an average higher rating than those items correlated to "leader's competence". >>> You can do this by looking at the mean responses to the ITEMS making up the factor, but not the factors themselves, and you might want to weight the component questions according to the results of the factor analysis (the loadings, I believe, but terminology varies across programs and books - not to mention languages!) but I am not sure you need to get so fancy - the simple mean may do nearly as well. HTH Peter Peter L. Flom, PhD Assistant Director, Statistics and Data Analysis Core Center for Drug Use and HIV Research National Development and Research Institutes 71 W. 23rd St www.peterflom.com New York, NY 10010 (212) 845-4485 (voice) (917) 438-0894 (fax) . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
