I have a study with the following data and I would like to calculate if there is or isn't any statistically significant difference between the 4 groups:

Test  Patients    Negative  NoDimer PotNeg  Total
W     317 (79.3)  12 (3.0)  194     19      31 (7.75)
WS    271 (67.8)  11 (2.75) 155     15      26 (6.5)
G     245 (61.3)  11 (2.75) 136     14      25 (6.25)
GS    260 (65.0)  13 (3.25) 144     16      29 (7.25)

I have 4 decision rules (W..GS) and 400 patients. 2nd column shows how many patients have negative decision rule (ant the percentage). 3rd column shows how many patients had a Dimer test negative (out of 2nd column patients) and the percentage (out of all patients). 4th colum shows how many of the patients in 2nd column did not have a Dimer test and the 5th column shows a possible number of patients with negative Dimer (out of 4th column) if they would have had the test done (determined using the proportion neg/pos from the patients that had it done). Last column is the sum of known and potential negative and proportion of all patients.

Is it possible to determine from these data if there are significant differences between the 4 decision rules. What methods can I use for this and do I need to use more data? If there is a method an example (or pointer) in R would be greatly appreciated.

Thank you.

  • $\begingroup$ Sorry I'm not sure I understand the data - what are decision rules? Are you just trying to see if the data in row 1 is significantly different from row 2, etc? $\endgroup$
    – RickyB
    Apr 5, 2013 at 0:44
  • $\begingroup$ Yes, I just want to know if the data in any of the rows is or not significantly different from the other rows. Decision rule means actually a clinical decision rule - based on different patient characteristics each decision rule calculates a score and a likelihood that a patient has or not a diagnostic. The first column is the name of the rule and the second one how many patients were classified as 'unlikely' to have the diagnosis $\endgroup$
    – rslite
    Apr 5, 2013 at 0:58
  • $\begingroup$ Perhaps discriminant analysis is what you're looking for? psychstat.missouristate.edu/multibook/mlt03.htm $\endgroup$
    – RickyB
    Apr 5, 2013 at 1:02
  • $\begingroup$ I took a look at the discriminant analysis, but I don't understand how it can help in this case. I just want to know if I can replace those 4 clinical rules with each other without a significant change in the results. Can this be done? $\endgroup$
    – rslite
    Apr 6, 2013 at 3:07
  • $\begingroup$ The format of your data is not totally clear, but try logistic regression! $\endgroup$ Jun 22 at 23:48


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.