to any maths people out there:

I'm doing a small study as part of my coursework, and I'm now writing up the results. Problem is, I am terrible at statistics and I have no idea what stats tests (like confidence intervals or whatever) are appropriate for my data.

Basically, the study looks at the proportion of asthma patients that have received stop-smoking treatment in the last year so I have a load of proportions representing how many patients stopped smoking out of the total number of patients.

e.g. This gives 28 patients out of 184 (15.2%) with asthma who are recorded as actually stopping smoking in 2009

Any stats geniuses out there? Can e-mail more details if needed.

Your not a real mathematician if you do Statistics!! Sorry i cant really help with Stats im terrible at it,could barely do it A level.

So I guess you want to calculate some measure of association like a relative risk (or an odds ratio, or a hazard ratio depending on how your study was designed), and then confidence intervals for that association.

Is it a cross-sectional study (i.e. you collected all the data at one point in time, with no follow-up), or did you collect the data on who had been given smoking cessation advice at the beginning, and then follow up those same people a year later (cohort/ longitudinal)?

You can send me more info if you want me to have a look, or we can have a chat at the wall if you like?

Thats not maths at all, I failed my stats module 3 times neil so I am not your man. If there are any real mathematicians out there can you help me with my pde's problem sheet?!

_________________
Unlevel-headed member of team Gadd-vey

Cheif Scout

well, its an audit, so the aim of the study was simply to look at the number of patients taking treatment in 2009 and I got the data by just doing pt records searches for all registered pts with smoking cessation prescriptions in '09.

I guess that makes it a cross-sectional study, since I'm not comparing the data to an older study or doing a prospective follow-up.

I'm wondering whether I need to do anything to the data, to be honest, since its not within the remit of audits to come up with association measures as far as I'm aware. I know I would if I was comparing, say, the uptake in COPD pts compared to the rest of the population but that would make it a case-control study and not an audit.

I just don't want some poxy examiner to ask "And why didn't you apply Spearmint's Log TreeBranch Polomint co-efficient's interval tests to this data?"

In order to transform your data, convert it to a percentage and you can arcsin transform it, giving you a nominal scale (i think)

Well I guess you need to decide whether you need to do any analysis at all, or just describe the data you have. If you have two groups you want to compare, or compare your group to a reference/ expected standard, you could do some simple analysis.

In terms of analysis, you could use a chi-squared test to compare proportions between two or more independent groups, if you want to see if there is a difference between the groups that is unlikely to be due to chance. If it is a 2 x 2 table with n less than 20 then you should do a Fisher's exact test.

This paper is quite useful if you want to choose a test.
http://www.sciencedirect.com/science?_o ... 1572c60e6f

awesome. that paper's well helpful thanks.

Neil, there are also lots of books in the library....and many have medical (type) examples. Not so helpful for me, but might be for you.