Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 2 down vote favorite

I am planning a study to collect data of patients dying after brain infection versus number of cases for the next 5 years in 2 cities where they manage the disease with different drugs.

  1. Will a year wise t-test be better or year wise Relative Risk of death and meta analysis be better to compare the results of the study?
  2. I am a beginner in statistics. A word of explanation would benefit me.

The data would look like

Year     City 1 Deaths    City 1 cases  City 2 Deaths   City 2 cases
2011     237                1124             10            1226
2012     228                1030             26           1181
2013     1500              6061             10           1122
2014     528                2320             32           1173
2015     645                3024             11           1232
share|improve this question
3  
I am not sure to figure out what your data look like. Maybe it would be benefit to make it clearer. However, if your variable of interest is of the type "time-to-event", then an analysis that takes time into account is definitely more informative than an analysis that only considers the number of cases. One reason for this is that censoring often occurs in such a study and a "time-to-event analysis" is especially designed for that setting... – ocram Mar 14 '11 at 18:38
1  
In addtion to @marco 's point, I think what you are looking at is a time-series and the independence assumption for observations of t-test may not hold. – suncoolsu Mar 14 '11 at 20:32
1  
I meant the data on mortality without any treatment, or with treatment when the same drug was used in both cities. The reason I ask is that with just the current data you cannot distinguish between two possibilities. One possibility is that the drug in city 2 is better, another possibility is that the drug is no better, but something else causes the patients to do better -- maybe weather or overall hospital quality or something else. – SheldonCooper Mar 15 '11 at 2:01
1  
If you can just claim that all other factors are the same, then it's fine to just do a t-test year-by-year or on all 5 years together. This is OK if this is just homework. If this is a real study and real data, then keep in mind that a lot of people will refuse to just believe that all other factors are the same. You will have to convince them somehow. In your previous question you had some pre-intervention data to convince them. Do you have something similar here? – SheldonCooper Mar 15 '11 at 3:25
1  
@DrWho Meta analysis is for combining results from multiple independent studies that have already been performed and published. This is not such a situation. (That's why I originally removed the meta-analysis tag from your question, because it will only mislead people who truly are concerned about meta analysis.) – whuber♦ Mar 15 '11 at 15:10
show 7 more comments

1 Answer

up vote 1 down vote accepted

Since your data are categorical, a t test is likely inappropriate for any parameter I can think you might want to do inference on. Here is a brief intro to a very similar problem as yours:

http://walkerbioscience.com/word-files/Categorical%20data.doc

share|improve this answer
Thank you very much. I will go though it and get back. i will have to search for the power point file mentioned in that document. – DrWho Mar 16 '11 at 4:30
Thank you very much for the simple but useful reference. – DrWho Mar 17 '11 at 5:40

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.