Rank drugs attributes based on their significance level - Need help Understanding Formula

Hi,

I would like to compare multiple values for one variable (i.e. several drug percentages for 1 ‘reason for choosing a drug’).

I've found the attached excel sheet set up with some formulas that are supposedly doing this analysis called the ‘Magic sig tester’ (attached). In the ‘dumping ground’ tab is where I can paste in our raw data, I then convert percentages to numbers and paste in the list of reasons and corresponding values either into the ‘Best of 3’ or ‘4 v ¬4’ tab depending on how many values (in this case drugs) we have for our set of variables.

The data I currently have in the sheet attached contains 4 drugs, so I have pasted the first 21 ‘reasons’ and the corresponding values into the ‘4 v ¬4’ tab.

Below the values are some formulas which use the overall bases of the drugs (at the top of the sheet) and the values I have pasted in to analyse how big the significant differences are between drugs for each reason (supposedly).

I then use the ‘Max’ values (column to the right of this data) to organise the list of ‘reasons for choosing’ into which ones produce the biggest differences between drugs.

Does anyone know what type of sig test is the formula doing? Does anyone know if there is a more intuitive way to rank these attributes?

Sorry if that’s not very clear, let me know if you need anything else to send them.

Mils

I'm afraid I am not enough of a statistician to recognize the particular significance test that is being applied here, nor can I recommend a specific test for this.

I might suggest you take the Excel formula and try to wade through the parentheses and references and get the equation written in a more algebraic format. That might make it more recognizable to a statistician. You might be able to then pass that formula through a list of common statistical tests (like this one from Wikipedia http://en.wikipedia.org/wiki/List_of...tistical_tests) and see which one matches up.

Considering the type of data, I might expect some variation on ANOVA or an F-test or a t-test or a Tukey's test, but I really cannot say for sure.