Statistics Advice please

O

Hi guys.
Just wondering if anyone can point me to a good statistics website? I'm having difficulty interpreting results and need to determine if they are statistically significant.

Thanks

B

Hey, what package have you used to get your stats (i.e., SPSS minitab etc)or have you done them by hand? What statistic is it?

F

I think Bobby is right, we need more details. E.g. if you use SPSS you can look at their help section (tons better now than it used to be). There are also text books which explain SPSSmprocedures and include output which would help also.

I realise that you may not be using SPSS but I would imagine similar principles would apply to the package you are using.

H

Maybe osiris doesn't know what SPSS is and you've scared her/him off!

F

Sorry

O

ha ha

No, I'm not using SPSS!
I'm using Prism 4 by Graphpad.
I'm studying a blood value in 2 groups of people (patient vs controls)
The patient group demonstrates a "spike" in blood value at their time of presentation. The control group doesn't have a similar pattern. I was wondering whether to compare the means of both groups, and then the spike value vs the mean of controls.
Not too sure whether a paired t test would be the correct statistical approach.
Any suggestions?

C

t-test I'd say. And you can easily do that on prism

C

Saying that, please bear in mind that I'm rubbish at stats

B

Yup, its an unrelated t-test, really easy to do (although you need to check your data is parametric first, if not use the Wilcoxon Mann Whitney U statistic).

Not sure how you do it in your package, but its fairly basic. you then get a number (the t statistic) that you need to look up in a table (or its likely that your package will give it to you). Before you start you need to state your alpha level (i.e., 0.05, 0.001 etc) at which you decide to reject your null hypothesis should your experimental t exceed the t under the null.

That's probably not very clear, I'm blatently not a statistion!

B

...And I can't spell...

C

On prism, just choose analyse and then t-test or something it's dead easy.

O

Thanks guys! I'll try it out!

J

if it not bayesian i wont touch it lol

C

I did mention that I was rubbish at stats

B

Ahh, BanD...I wondered if this might tickle your fancy come on, tell us how it should be done, the gold standard way...I love it when you come over all stats...makes my heart flutter...

5657