talkmagic.co.uk

Hey - I'm writing up a practical psychological report, but need to include some stats for obvious reasons.

Having missed 2 lectures through illness a few weeks ago, a few things I am buggered about...so in a nut-shell, can any of you help with the following?

What on earth is a t-test? and what is the difference to the other tests? I'll need to know, because I think I have to include one

What is nominal data and what does it mean if a certain test require "higher than nominal data"?

and what is parametric and non-parametric?

Cheers in advance!

Righto, a trawl through my rotting mind recalls that a t-test shows how different the means of two sets are from each other despite times when you have to make assumptions about the data.

Nominal data is data that passes rudimentary tests for inclusion in a set. If you have two sets, male and female, you can nominally categorise each subject in a study unequivocally into those two groups, then count the number of members to get the total population, the distribution, etc. Another use is to describe the smallest statistically significant population that will give good predictive results.

Parametric data is derived. So the mean, standard dev, etc. of a population is all parametric data.

thankyou!

so is parametric data not something that falls into something like a bell-shaped curve as opposed to "random" distributions?

and what is "Higher than nominal data"?

Cheers boss!

I'm no statistician but here's what I think.

The t test relies on an underlying assumption that the data in the two sets that you're comparing are normally distributed (i.e. bell shaped curves). It also assumes that the variance (standard deviation squared) is approximately equal in the two data sets. In other words if one data set has a massive amount of variation and the other doesn't then the 't' test assumptions are not met. If these assumptions are not met then the results of a t test should be treated with caution and you shouldn't quote the 'p' value that tells you whether the two sets of data are statistically significantly different. Things like 't' tests and ANOVA are parametric tests because they both have these assumptions of normality and equal variance. Non parametric tests are for when the data is not normally distributed - there are a whole suite of non parametric tests available. If I'm wrong about any of this then I must apologise to Keith - our dept. stats advisor who patiently tries to embed this stuff in my tiny brain.

Hope this helps.