Hi all,

It has been a while since I formally studied and need some statistics advice about what statistics tests I'll use in my study. I have whittled things down but could do with asking someone some questions. Would anyone be able to help me by PM?

Hope it's OK to ask.

I think it all depends on what data you are analysing, as there are many equations and systems available. If you are simply comparing two datasets for significance levels, then the Mann Whitney U test is fairly straight forward. ANOVA (analysis of variance) is also a good system.

Thanks Ginga. I've a form to complete and it's more comprehensive than I would have liked and I have to provide a relatively detailed analysis.

Hi Delta:
It depends on your data; if your data is normally distributed - a Poisson curve - (as most "ideal" data tends to be) then you'll be looking at the following:
student's t-test (two independent variables)
ANOVA (multiple indepedent variables)

Mann Whitney is good if your data is skewed to the left or right (so not normally distributed).

Best thing to do is get some data, plot a histogram of it and then find the in-house statistician to give you a hand.

Get the Andy Field book: Discovering Statistics with SPSS. It'll take you through the very basics, assuming no knowledge of statistics, right up to the level of the decisions you need to make when choosing statistical tests.

======= Date Modified 16 Mar 2009 11:31:10 =======
Thanks all. I've to give full details about my analysis before I even data collect. This is what I'm struggling with. I can follow what you are all saying but this form - yikes!

hi delta i am also going into statistics. I don't know how your data is like. But here's what I have been reading so far:
By Jae-On Kim & Charles W.Mueller: Introduction to Factor Analysis: what it is and how to do it (I like this, its simple)
By Derek Rowntree: Statistics without tears (this one is a good refresher)
By Jeremy Miles and Mark Shevlin: Applying regression & corellation: a guide for students and researchers (its really good)
I know we can always ask for help from our stats buddies. But if nobody is around, at least these books are here..:-)

The easiest way to go about it is probably to just modify the statistical methods sections of research papers that will employ a similar methodology to yours. There are also many formulae and sample size calculators available online if you do not yet know how big your study will be.

Thanks all. I think I'll just need to take the time and apply myself to relearning some stuff but some good suggestions and so thank you.

Sorry- but most of this information is incorrect. There are more statistical tests than just 'comparing groups' i.e. ANOVA/Mann-Whitney/T-tests...

The most important thing with getting the right statical test is to check your assumptions. So, ANOVA/t-test's assume that the data form a normal distribution - a bell curve- (not a poisson curve) and that the group standard deviations are equal.

Mann-Whitney U test can be used on non-normal data but still assumes that the group variances are equal and that the distributions between groups look the same.

You may also want to create a model for your data to look at risk factors, or predict outcomes, look at relationships....etc

Anways, I could go on forever explaining the various statistical methods of analysis! If the OP would like to PM me with a description of their data and their hypotheses i.e. what they want to prove/disprove from the data, I can recommend methods of analysis etc.

Billys

P.S. I am a second year PhD student studying statistics, I have an undergrad in stats and do a lot of consultancy work so do know quite a bit about this!

PM sent, you wonderfully kind person!

Not a problem have PM'd you back!

You've been great and saved me some stress, many thanks!