Help with modified Poisson regression, risk ratios for medical study

[James Hearn]

Hello all,

I am a research associate at a Canadian university. I am brand new to STATA, just purchased a license today, I previously only used SPSS. So I am brand new to the program.


We are writing up a paper of our study, and I am in need of assistance. The study primary outcome for our study is surgical site infections in patients in two groups, before our intervention (2014), and after our intervention (2015). I would like to perform a regression, and look at the relative risks (both crude and adjusted) and 95% CI's of surgical site infection when controlled for various factors.

The factors I would like to look at are exactly the same as this image (attached), which is a previous chart using abbreviated data. Would it be possible to have some assistance on how to perform this analysis in stata? My data is already completely coded in excel.

I would be hugely appreciative in any assistance.

[Clyde Schechter]

If you are brand new to Stata, my best advice would be to get acquainted with the program before attempting to tackle a complex problem with it. The first step is to install it on your computer. Instructions came with your package or download and you should just follow those.

Then, launch Stata, and click on the Help menu and select PDF documentation. (You need Acrobat Reader or some other PDF reader--if you don't already have one, do install it. You will need to refer to the PDF documentation often while learning Stata, and from time to time once you are familiar with it.) Read from beginning to end the Getting Started [GS] and User's Guide [U] sections of the documentation. Pay particular attention to the [U] section called "Commands everyone should know" and click on the links to those manual sections. It will take you a while to do this, and there is enough there that you won't remember it all when you're done. But these two sections will expose you to the basic operations of Stata as a program and to the Stata command language. You will have seen all of the fundamental commands that you need to get commonly needed things done. Then when you actually start to use Stata, you will have some sense of what commands might be relevant and you can go back to the documentation again to refresh yourself on the details. The time you invest in doing this will be amply repaid once you get to actual work.

You will find that the Stata language is rather different from SPSS, and that the Stata model of data management is also different from that which works best in SPSS. So it will be a bit of an adjustment for you. All the more reason to become familiar with the documentation first; your instincts developed from using SPSS will sometimes be useful, but sometimes will lead you astray.

One command you will need to learn for your specific problem here, that is not, if memory serves, covered in [GS] or [U] is -import excel-, which is found in the Data Management [D] section of the documentation.

At the risk of being cynical, in my experience as an epidemiologist, it is rare to encounter a data set in Excel that doesn't contain numerous errors and problems. So before you plunge ahead with Poisson regression, be sure to carefully validate and clean the data in Stata.

Once you get started on your project, have a Stata data set in hand, and begin to setup and execute your Poisson regression, do post back with specific questions as they arise, and people here will jump in to help. To maximize your chance of getting prompt and helpful responses, read the Forum FAQ and follow the very helpful advice there about how to ask questions, how to post data examples, and how to post Stata code and output.

[James Hearn]

Thank you for the post Clyde ! I appreciate you advice. I am in somewhat of a predicament though as I need this completed by Monday. I was only asked by my supervisor today to try and get it in time for her to give a presentation.

If there is anyone reading this who is able to do this modified poisson regression, and is willing to do this for a fee, please let me know, I would love to hire you for it and would be very grateful.

James