I find this question really interesting and some of the answers even more
so.
Jeremy: how much would you say that the shape of the distributions matter
at all in this case, given what the central limit theoram implies about
using parametric tests, even with only 50 or so participants
http://www.graphpad.com/www/book/choose.htm
One other thing: has anyone talked about transforming the data
statistically e.g. log transformation? This often doesn't work for me too
well with small distributions, but is always worth a try.
May I throw in a related question? Has anyone ever tried a Box-Cox
transformation for really skewed distributions? If so, did it work and do
you know of user-friendly free software that can do this?
Significantly skewed items are not a problem.
Very, very skewed items are a problem.
If your items are not significantly skewed, collect more data and they
will be. I don't think any program except SPSS even calculates the SE
of the skew, it's such a useless statistic. Streiner and Norman, in
the book "Health Measurement Scales" suggest that if 80%+ are
responding at one end of the scale, you have a problem, otherwise, it
doesn't matter.
jeremy
On 10 November 2011 07:19, Sandi Dheensa <[log in to unmask]> wrote:
> Thank you all for your quick responses! :) In answer to your question
> Jeremy, yes I do mean that the skew is statistically significant. Thank
> you for pointing that out...(what a rookie mistake for me to make,
> hahah). I'm going to have a go at shifting all of the response options,
> like in the example about the doctor, readministering the questionnaire,
> and then rerunning the analysis...I think this is a legitimate move but
> if anyone else has any words of advice, please let me know! If the items
> are still skewed once I've rewritten them, you'll hear from me again
> (either by email or simply because the sound of my weeping will be
> audible for miles) Sandi
>
>
>
>
>
>
>
> ________________________________
> From: Dan Holman <[log in to unmask]>
> To: [log in to unmask]
> Sent: Thursday, 10 November 2011, 14:56
> Subject: Re: Scale Construction
>
> Do you have examples of some of the question wordings? Maybe we could
> think of ways to re word them to get a wider range of answers. Dan
>
>
>
|