Hello:
I am planning on running either a Chi^2 or z-test for population
proportions. Is there any way in SPSS to control for multiple tests? For
example, Sheffe, Bonferroni correction, etc. might be appropriate, but
having reviewed the statistics literature, the Bonferroni technique has
not ( I don't think) been applied to this type of data (binomial, observed-
expected frequency). Bonferroni divides the error rate (e.g., .05) by the
number of tests made to get an error rate for each test that overall does
not exceed the familywise error rate. Are you aware of any other
transformations/techniques that specifically address the binomial observed-
expected frequency multiple test issue?
My data includes the results of a collusion screening technique (no
collusion vs. colluion) by postal code. I have 325 unique postal codes,
and when I screen out samples of <5 (for both flags), I have about 40
comparisons left to analyze. The expected proportion (random chance) is
87.5% no collusion and 12.5% collusion for each.
Any suggestions most welcome and appreciated.
Thanks
Scott Graham
Doctoral Student
University of Nottingham
[hidden email]
[hidden email]
http://www.nottingham.ac.uk/business/lixsg10.html=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
