MAY DAY MAY DAY looking for stats guru
I'm suddenly confused:
If my sample has 68 men but only 8 women, is the data considered skewed and how do I transform it?
Or can I go ahead and use this data.

satchi satchi

Hey Satchi- I am not a stats guru by any means, but I will try to answer! It depends really whether gender has any relevance to what you are looking at....if not, then the difference won't matter. For example, we have always been told that if gender isn't relevant to what we are studying, then we don't need to even mention how many of our participants were of each gender in our method/participants section. As far as I am aware, this isn't what is meant by skewed data (this is more to do with a skewed distribution curve etc, i.e. the data doesn't meet the criteria for normality), and I don't think a transformation will be relevant here. As long as you are looking at something that is not gender specific then you should be able to use the data you have. Obviously there is unlikely to be enough data to make comparisons between genders, and you will might have to take the gender ratio into consideration when trying to generalise your results, but apart from that you should be okay! I always try to get a fairly even mix for my studies, but sometimes it's just not possible so don't stress about it! KB

hi keenbean
thanks
I really needed to read your post
satchi (up)

Hi Satchi,

as Keenbeen is indicating it all depends on whether the gender may have any impact on the outcomes, and as such whether there could be bias.

I think that if you are doing a qualitative study having a balanced sample is often not that important, as you are not looking at a representative sample of society. Evenso it may be worthwile to mention that you are aware of the relatively large amount of men, and possibly explain / justify why this.

However, if you are doing a quantitative study the sample seems relatively small and I am not sure whether it would provide statistically significant results (as evidence that results are not based on chance). With this type of study you probably do have to assure that there is no bias (unless, as mentioned, gender is irrelevant). In my opinion the only way you can do this is by making the sample bigger, assuring that you have more female participants.

hi Rick
yes you are right, gender in my case did not show any significant effect.
I'm also not able to increase the number of women because the study is over.
thanks very much
satchi

Hey again Satchi! Just so that you don't panic over participant numbers, 76 will be fine for some studies- it depends on what you're doing. The power calculation for my study showed that I only needed 30 participants in each group for 80% power at a 5% significance level, and with 60 sets of data I already have a plentitude of significant results! Of course some might require more than that, it depends on the strength of relationship you are expecting etc. Good luck with the analysis, KB

keen bean where did u do the power analysis?
did you use an online calculator?
thanks
satchi

======= Date Modified 27 Jan 2010 15:10:07 =======
Hey Satchi, yeah I used an online power calculator, the one at http://www.danielsoper.com/statcalc/ and also the one at http://www.surveysystem.com/sscalc.htm I was using it for a multiple regression sample size calculation, I'm not sure if it does sample sizes for every different kind of analysis, but if you google 'online power/sample size calculator' then you will find a whole load of different ones so you should find one to suit! KB