What kind of Excel test or data analysis to use in a statistical study? Example given?

Question

I want to find out if there's a correlation between a number of events (column one) noted 1, 2 , 3, 4 (as took place only once, twice etc) and a certain consequence (let us say death) having the number of subjects that went through that events (once or twice etc) as the second column and the number of subjects that suffered the consequence as column 3. What test to use?

first column: number of interventios (surgery) that patients went through: values are 1, 2, 3, 4
second column: number of patients that went through those interventions: values are 27, 9, 3, 1
third column: how many died after they underwent respectively 1, 2, 3, or 4 interventions: values are 9, 4, 2, 0. So the one guy that went through 4 interventions did not die, 27 went through 1 intervention and out of these 9 died, etc.
All I want is to find what corelation is between the number of interventions and the mortality. I hope I was clear.
I have that analysis installed, and I have no ideea how to recess something ... :D

Answer 1

add a fourth column and make it percent of those effected (ie if the sample size in column 2 is 100 and 50 people went through the event, the fourth column would be .5). Then regress column 4 as your dependant variable against column 1 and check your correlation coefficient

You need to add in the data analysis pack in add ins...

Minitab would be a better tool

Answer 2

I think the function is =correl(array1,array2) but you can look up in help to get the function name.

This function will give you the p value from -1 to 1, where the negative number indicates a negative correlation (if one thing happens something is likely not to happen) and so on. Depending on how much data you have will help you figure out what the strength of your correlation might be.

For example if you're looking at deaths by paper clip, you should realize it is a pretty rare event. You'll need lots of paper clip users. Then you can look and see if "death by paperclip" is correlated with the paper clip design (the long roundish ones, or the butterfly variety) or if in use.

If on the otherhand you want to look at the correlation of white shirt to spilling coffee in the morning, you'll only need a small sample (couple hundred).

If you're looking at a direct impact, then you need fewer samples still.

Answer 3

You should do some form of model. A correlation test is too simplistic and will miss non-linear relations. Often you need higher dosages or number of events for any consequences to be noticeable.

Some form of GLM would be appropriate. Though i dont know how one would do such a thing in excel.

It would help to know what the data is about or to see a prelim analysis (i.e., plots) to help with determining what is an appropriate course of action.

Answer 4

you're particularly a lot there! you've were given your 3 columns set up. enable's say they are in columns A, B, and C, rows one million-4. Create a fourth column, D. In D1, enter =c1/b1. it truly is the mortality fee from the first surgical treatment. replica and paste down, to locate the mortality prices from the 2d, third, and fourth surgical treatment. Your mortality fee from surgical treatment #4 should be 0, because no one died on the fourth surgical treatment. Your Quote: "All i choose is to locate what corelation is between the type of interventions and the mortality fee. " solid! those parts are in columns A and D. This formulation, even as typed in cellular f3 will do properly. =Correl(A1:A4,D1:D4) ought to get you your correlation. wish this facilitates!

Answer 5

You can do simple hazard stats, but your model should be survival analysis...not sure this is possible in Excel.