How to organize a dataset with many attributes

Question

I have a set of data which consists of many different types (measurable, categorical) For example: name measurable_attribute_1 categorical_attribute_1 measurable_attribute_2 categorical_attribute_2 ...

Number of attributes may grows quite quickly during my study: into my spreadsheet, I can as many new entries as attribute... I have about a hundred of entries in this classification scheme, about 70 attribute, so far, and I am at the beginning of my data collection.

I would like to perform statistical analysis of this data set. For example, what are the common features of the entries that have a similar categorical_attribute and this range of values of measurable_attribute.

Well, I would like to generate relationships between attributes in order to create training images. However, I am not sure of how to organize the data prior to classification. Even though, should I organize the data?... (referring to this question)

Also, I can hardly gather entries into classes.

I do not want to introduce any bias obviously.

I am also quite new to statistical analysis (but eager to learn).

Answer 1

You could look at Exploratory Factor Analysis. It will tell you which attributes are the most similar to each other.

Answer 2

One option would involve using optimal scaling principal components analysis. The approach allows you to state your measurement assumptions about each variable (e.g., nominal, ordinal, numeric). I've used it in SPSS: see the Categories Add-On module (i.e., Analyze - Dimension Reduction - Optimal Scaling). I'm not sure, but the homals package in R may also implement the procedure.

A quick Google ( http://www.google.com/search?sourceid=chrome&ie=UTF-8&q=optimal+scaling+principal+components) revealed this reference: http://takane.brinkster.net/Yoshio/p009.pdf