MegaSack DRAW - This year's winner is user - rgwb
We will be in touch
I have this question to answer as part of an assessment that I need to do.
What descriptive and inferential statistics would be appropriate to analyse quantitative data to show that real rather than chance differences existed between two groups? Describe how the analysis would be carried out and what assumptions have to be met for the analysis to be valid and reliable.
Now I know what I think the answer is, but it's a long long time since I did any stats.
Please help me.
student t and f tests on covariance/correlations/anova?
The question is far too open ended. Just about any statitics analysis can be used to test a difference between two groups.
The question doesn't ask which moment is of interest. Generally it would be the mean, but is could be anything.
At a guess the person who set the question wants you to answer t-test and you have a continous variable and a large sample.
An ANOVA would probably do the trick, the p-value obtained would tell you whether you have statistically significant test. Hard to know exactly without more detail on the dataset though.
Most statistical tests assume that the data is normally distributed and the variances of different groups are equal, some have further assumptions such as random sampling, but it depends on the test you're using.
Those answer are pretty much what I was thinking. I agree that the question is a bit rubbish. Am pretty tempted to tell them that their question is rubbish and sacrifice a few marks. We have a 100 word limit on this question too.
Am pretty tempted to tell them that their question is rubbish and sacrifice a few marks
and there are lots of motor cyclists in hospital who were "right" too...
No point being right if it costs you dear.
If they've given you 100 words for a v general question, just play bingo with the keywords and a very very basic praecis of your methodology. If they wanted a dissertation on comparative statistics they would have made it clear thats what they wanted.
I think its actually quite a good question...
I see the question as less about what actual statistical test you use but how you arrive at the reasoning to use that over anything else. So you could imagine that the two groups are continuous, normally distributed, equal variance data sets and a standard normality test ANOVA/ post-hoc will do. It could be that the data set has outliers/leverage points/ unequal variance; its up to you and how much you want to impress the examiner.
[url= http://www.whichtest.info/index.htm ]here[/url]
If in doubt chi squared / fisher's exact test.
One of those Google questionnaires 🙂
I think its actually quite a good question...
Yes. It's testing that you actually understand why you do certain tests over others. And the limitations of those tests. Rather than just churning out buzzwords and the formulae you learnt the night before the exam.
So you could imagine that the two groups are continuous, normally distributed, equal variance data sets and a standard normality test ANOVA/ post-hoc will do.
I'm afraid it wouldn't, you'd have to make many other assumptions, and what more is ANOVA and T-test only test differences in means, not, as the question asks
which could refer to any difference, not just differences in means.show that real rather than chance differences existed between two groups
I'm afraid it wouldn't, you'd have to make many other assumptions, and what more is ANOVA and T-test only test differences in means, not, as the question asks
Opps! Which assumptions did I miss?
which could refer to any difference, not just differences in means.
Err not sure what you mean?
Just about every statistical test tests for a difference in the means of two data sets, you dont really test for anything else, hence why one of the main assumptions of an ANOVA is that the data is normally distributed. The two main assumptions of an ANOVA are homogeneity of variance across the sample groups and normally distributed data, any other assumptions would be a lot less important than those.
Just about every statistical test tests for a difference in the means of two data sets, you dont really test for anything else, hence why one of the main assumptions of an ANOVA is that the data is normally distributed
Well, the f-test looks for differences in distribution, but ultimately all these test are probabilistic, and none can show real differences beyond chance. The only way to show real differences is to use Population rather than Sample data. Taking that into account, the best way to show what you want is to look at effect size, i.e. significance rather than statistical significance, which is often obscured by large populations