I wonder why statistics professors seem to ignore such topics while professors in education, psychology, sociology, and other fields enthusiastically use them. Maybe I am very wrong in my perception.
They are probably not used as much in other disciplines so it is just a matter of priority, they cover the most popular topics first so it maybe a time factor
Hi Raid,
coming from a background in quantitative linguistics (rather than mathematical statistics), my impression (at least from the literature) is the same as yours, and that there is some truth to the skepticism toward factor analysis, at least. Venables and Ripley, in their book "Modern Applied Statistics with S" say something along the lines that it is hard to find cases in the literature where the factor analysis model (and its assumptions) fits the data well. I think Rencher expresses similar views about factor analysis in his book "Methods of Multivariate Analysis". I don't know how widely shared these sentiments are among statisticians, but I note that these are widely cited books.
Factor analysis has been used in my own field, corpus linguistics, but in my paper "Multivariate Analyses of Affix Productivity in Translated English" (available from my RG profile) we discuss why factor analysis probably isn't the best choice for that kind of data, and discuss alternative multivariate techniques.
My highly subjective impression is that data analysis in specific disciplines are sometimes caught in some "canonical" technique (one example being factor analysis for multivariate corpus linguistics, established by Doug Biber's work). The "canonical" technique then becomes entrenched as THE technique even if mathematical statisticians are skeptical, since there is (often?) scant communication between the fields.
Hello Gard,
Thank you for your input. You have shared with us very useful information above.
When I teach principal components analysis, it is straightforward to explain a linear combination in the variables (X's), but when it comes to factor analysis, I need to explain why we can assume that each X can be written as a linear combination of some "factors" that may or may not actually exist. The only reason why I use factor analysis is that when I was a graduate student, during one of the summer terms there was no other course offered that I had not had before, except Factor Analysis! There were maybe 27 Psychology students and 3 Statistics students in that class. I suddenyly became "one of the very few" students who knew how to use factor analysis (and other methodologies) with data from the social sciences, which we needed in many consulting projects that statistics students were involved in.
As long as factor analysis is actually "correct" to use, it is good to know about such methodologies.
Because FA techniques requires higher skills and essentially a lot of "imagination" to identify susceptible research problems and how to solve...look in Data Reduction chapter (include at least Discriminant Analysis and FA), and then you can glimpse to the potential of this technique about how to convert anyone natural repetitive event in nature (from grief to stock market) to identify the most important variables to "build" a current or predictive mathematical model and finally (if you wish) transform this new data into score system or similar, but you have to make a lot of reflection about this technique practical applications.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling