#### How do I relate two variables?

M

Hi,
I've not attended stats training yet so any help with this data analysis/stats query is much appriciated. I've been told to play about with my data (just started collecting, so don't have much of it yet) to recognise if there are any patterns/correlations.

I've processed one half of the data (results of a clinical test) and I would like to relate it to results from a questionnaire, but I don't know how where to start. I don't know how to process my questionnaire data, I don't know how to relate the two. I'm working just on Excel at the moment. Most of the data is on SPSS but I'm finding it hard to get my head round SPSS without the training. How can I relate these two variables?

Many thanks

D

Hi mittow,

well done for collecting your first data!

I would recommend that you start reading a bit about statistics. My personal favourite for beginners is "Discovering Statistics with SPSS" by Andy Field. It is a bit bulky, but it is easy and enjoyable to read.

The first thing to do with the data is to run some diagnostics: is the data normally distributed? Are the observations independent? Do you have paired data (one participant is a control for another participant) ? Do you take repeated measurements? Is your outcome(s) binary, categorical (ordered or unordered) or continuous? All these questions will affect the tests you should chose.

After I know my data, I do some simple plots to visualise my data. I move on employing with some simple ranking tests, testing the null hypothesis. After that, I move to more complex statistics. After you know what you are doing, it is a good idea to have a chat with a statistician.

H

Quote From mittow:
Hi,
How can I relate these two variables?

Worth noting that a questionnaire is not a 'variable', unless it only asked a single question. Rather, a typical questionnaire will have generated a number of variables. How you relate these to the clinical test results will depend on whether the questionnaire questions have answers that are continuous ("How tall are you in cm?"), categorical ("What is your smoking status: current, ex or never?") or open ended ("Describe your symptoms"). The testing you do may also depend on the number of participants (i.e. sample size) and as DrJeckyll says, the distribution.

Do you have a specific hypothesis at the outset? Be aware that if you do statistical tests on a large number of variables, you may find a correlation by chance, which isn't actually meaningful or cannot be replicated ( http://en.wikipedia.org/wiki/Multiple_comparisons_problem ).

While you are waiting for your training, ask around your department to see if there is a medical/bio-statistician you could have a chat to about your study, before you collect any further data.

Good luck

M

Quote From HazyJane:
Quote From mittow:
Hi,
How can I relate these two variables?

Worth noting that a questionnaire is not a 'variable', unless it only asked a single question. Rather, a typical questionnaire will have generated a number of variables. How you relate these to the clinical test results will depend on whether the questionnaire questions have answers that are continuous ("How tall are you in cm?"), categorical ("What is your smoking status: current, ex or never?") or open ended ("Describe your symptoms"). The testing you do may also depend on the number of participants (i.e. sample size) and as DrJeckyll says, the distribution.

Do you have a specific hypothesis at the outset? Be aware that if you do statistical tests on a large number of variables, you may find a correlation by chance, which isn't actually meaningful or cannot be replicated ( http://en.wikipedia.org/wiki/Multiple_comparisons_problem ).

While you are waiting for your training, ask around your department to see if there is a medical/bio-statistician you could have a chat to about your study, before you collect any further data.

Good luck