What is/are Latent Dimension(s)

hlsmith

Less is more. Stay pure. Stay poor.
#1
In the following description of Discriminant Analyses, what is meant by "latent dimensions"?

"It is a multivariate technique that considers the latent dimensions in the independent variables for predicting group membership in the categorical dependent variable."

Is it related Type III effects and/or matrix algebra?
 

Jake

Cookie Scientist
#2
Latent dimensions/latent variables are variables which we do not directly observe, but which we assume to exist (in at least some instrumental sense) in order to explain patterns of variation in observed or manifest variables. For example, in personality research we might give someone a questionnaire that consists of, say, 20 questions about their personal relationships, social life, etc. At the most concrete level, this gives us scores for each person on 20 separate dimensions. However, we typically assume that there is, say, a latent dimension of "introversion/extroversion" that can account for much of the systematic variation that we see on the 20 observed dimensions. So with some statistical assumptions we end up deriving from the scores on these 20 observed dimensions a single measurement on a latent dimension.
 

noetsi

No cake for spunky
#4
No a confouding variable is normally an observable variable not in your model which explains the results you found. Or in some cases it may be two variables who's effects can not be seperated from each other so you don't know which causes the results.

A latent variable is an unobserved factor which is measured by a series of variables you can observe. Intelligence is the classical example, we can't directly observe it, so we try to measure it by a series of observed behaviors we can see.
 
#5
If you want to find out more, look up "latent trait theory". Commonly assumed latent traits are things like intelligence, literacy, creativity, etc.

John