Identifying the largest determining factor

[SanDiegoMike]

Identifying the "largest determining factor"

Hello,

This is not strictly a statistics forum, but I'm hoping you guys may have sufficient background to help me out. I'm an engineer by trade, so my stats background is poor and I have had not had much luck searching for the answer or asking colleagues.

I have a database, which is similar to the following example in which we list a collection of 'coins' of various values. These coins all have different characteristics, ie: year, color, size, shape, and weight. I would like to determine to what degree each of those characteristics are most likely to determine the coin's value. Or at the very least, determine which of the characteristics is most predominant. My searching has led me to analysis which requires some form of functional relationship between say 'shape' and 'value' such that correlation can be determined, but I don't know how I would convert shape (ie: circle, square, octagonal) into a variable. My colleagues have suggested scatter plots to identify relationships, but my data sets are huge, and I would prefer something with a mathematical foundation.

If anyone could point me in the correct direction with regards to the appropriate analysis methodology, that would be fantastic.

thanks,
-mike.

 

[bpet]

My colleagues have suggested scatter plots to identify relationships, but my data sets are huge, and I would prefer something with a mathematical foundation.

Data exploration is arguably the most important step of any data analysis methodology so I wouldn't discount visual tools just yet. Maybe start with a scatter plot matrix for the continuous variables and box-and-whisker plots for the categorical variables.

 

[alan2]

Try this reference for a start. You have categorical variables with more than two levels which can not be ranked (as in shape).

http://www.psychstat.missouristate.edu/multibook/mlt08m.html

See the example on faculty salary. Also try Sheskin's handbook.