Omar asked:
------------------------------
I have installed the 'lambda' module to obtain a Goodman Kruskal measure of
association between nominal variables. Can anyone tell me what is the
maximum number of rows and columns that can be used and secondly how one
interprets the result? I obtain lambda_a, lambda_b, and a plain lambda.
What do they each mean? Thanks, Omar
Omar,
I've not seen lambda used for years. It is a proportional reduction in error
measure. Lambda a and b are asymmetric. For example, you compare the number
of miss classifications using the row (or column) marginal mode to the miss
classifications using the modal cell in each row (column). Soooo, a lambda
of .20 means you have proportionally reduced error (misclassifications) by
20% when you know the category the person is on the independent variable and
predict the mode for that category. The symmetric lambda must be a weighted
average of the two lambdas.
If a marginal distribution is highly skewed, then lambda will often be zero
or close to it, because knowing the other variable doesn't help. This is
because it only uses the mode for prediction and misses important
differences in the distributions.
Alan Acock
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
