I felt it was a pecuilar paper, although I did appreciate that it made you think about the techniques used rather than repeating methods blindly.
The first question threw me off to begin with (the laptop battery question) and I don't think I got full marks in that question. I used the Normal approximation, but realised after that I foolishly divided by the sample size twice, which clearly would have given a wrong answer.
I had only used the Fisher Transformation to test the correlation coefficient once while doing past paper questions, so I am happy that I spent what I thought would be wasted time knowing when to use it. My answer seemed reasonable, which is always a good thing!
I think I managed to just eek out the marks for the coin toss question in the last 15 minutes.
The one question which really got me was the testing of the binomial parameter, p (one of the last three questions). I used the method of moments to estimate p to be 0.5, and used the chi-squared test to test the distribution. However when I read the next part, which asked us to state the effect of instead being told p=0.5, as opposed to estimating it, I didn't consider that it would mean we could include one more degree of freedom, and we could then easily accept the null Hypothesis again based on the first. And then the next part told us to test p=0.5, which made me think that I'd used the method of moments incorrectly, so I changed the labelling of my initial chi-squared test to part (iv) instead of part (ii) *shakes my head*.
Last edited by a moderator: Apr 29, 2012