I've had to waste loads of time today grappling with the forces of reliability testing for a questionnaire system of measurement. Anyway, for demonstrating test-retest reliability, I finally settled on intraclass correlation co-efficients, Lin's Co-efficient of Concordance and a decided to go a bit descriptive with doing a Bland and Altman Limits of Agreement Plot. Now, I'm pretty sure the first two are fine - even though they are ordinal and strictly speaking...but everyone does it (papers galore use the same approach and I don't want to be left out with naff non-parametric tests that only null hypothesis fetishists use. But then there's the Bland and Altman plot - someone has said that it's naughty to use it on non-normally distributed data - and I've used convenience sampling and it's non-parametric. I was sweating at this point and thought about how I can justify it and there's a very strong argument for treating ordinal scales with 5 or more points as interval level data. And then a master of the psychometric dark arts has said that although strictly speaking it is naive to treat ordinal level data as interval (and so parametric), it can pragmatically be useful. And then people in the behavioural and social sciences do it. Sorry, I'll get to my point... So, I'm thinking I should go with using a Bland and Altman plot for agreement and not worry about the fact that I've used parametric statistics for ordinal level data, since it is popular and I can justify it. I have a pragmatic research world view anyway - and it's about the ends justifying the means. And I've spent a/wasted 2/3 of a day having to look into this. For anyone who knows about this stuff or is interested in this post, does that sound fair? Or is it a complete (turkey)?

I thought this was going to be about car insurance and was going to be able to put to use my knowledge of the industry.

I'm also a statistician but after two days of learning about Access, it's a bit much to understand, I'll have a look tomorrow if nobody else has replied.

I also thought it was going to be about insurance! Then again I'm not a statistician, and reading the message left me completely baffled :p

I like Bland-Altman plots. :)

Can't think why they can only be used on parametric data though. Then again there seems to be a lot of rules around stats...think I'll be breaking a lot of them in my thesis...

I'm doing a research on medical tourism. My population is medical tourists visiting hospitals in Town A & B. How could I determine my sample size for convenience sampling?