Is it sufficient to test at 5% always or do you get more points if you check at 1% as well? I am asking because I've seen on the ASET materials that 2 conditions were checked but I do not recall seeing it in the exam reports.
I think you'll be fine just checking at the 5% level (unless a question specifically directs you to use a different significance level). I can't recall anything other than that appearing in the examiners reports. Where ASET checks at other significance levels, I suspect this is mainly to draw out how extreme the observation is (eg. the null hypothesis would even be rejected at the 1% level).
One more question on the Standarised Deviations test this time. After doing the test do we need to formalize the conculssion by performing a chi squared test? The reason I am asking because if we look at the solution of September 2001 Exam Q 8, the sandardised test shows skewness of data and by eye we see the distribution is not approx normal. The solution by the Institue does not give an additonal formal confirmation of chi squared on the observed vs. expected proportions. When I did the chi squared test with 7 d.f at 5% (the Act Ed materials suggest doing it), I got the opposite conclussion - the wrong conculssion. Maybe it is not even required to do a Chi squared for standardised deviations test? Would I loose marks?
No I don't think you'd lose marks for not doing the chi-square test to formalise the results of the individual standardied deviations test. It's not done in the Examiners Report, which is the best guide we have as to what they were looking for. I'd rarely do the chi-squared part of the individual standardised deviations method - and in particular not here. The notes remind you that doing a chi squared test where the number of expected values in a group is less than 5 won't give very reliable results. Since you only have 9 age groups here, you'd have to group results into positive and negative to just get near to 5 expected values - so it reduces to comparing positives with negatives which already looks fine as it's 5 vs 4. The real problem when looking at the standardised deviations here is the two values greater than 2, which is highly unlikely in such a small sample, leading to the conclusion of rejection.
