I have a few little nit picky issues; it would be appreciated if anyone could assist
It is given in the S3 book that the bias of an estimator theta for a population parameter T is E(theta) - T
Yet I recall in one mark scheme (I can probably find it if my explanation here isn't clear enough), that the bias of such an estimator, whose value of E(theta) was to be worked at, was given as (+/-). How can this be; surely E(theta) - T will give a sign dependent on whether E(theta) is less then, equal to or greater than T? I didn't get the impression that either putting plus or minus was acceptable; there were no brackets or anything to suggest this, just the modulus of the bias with a +/- in front
Also, and I guess this is quite similar to the other recent questions on this thread, can I just put exact fractions for Expected data and ((Oi - Ei)^2)/Ei) or do they want me to just write down a few decimal places
Frankly it seems bizarre that they could deduct marks from you for just plugging it all (that is, the chi square statistic, I understand it might be a bit tricky to not first write down your Ei) in to your calculator instead of writing it all down which is pointlessly time consuming and reduces accuracy when you round things all the time
Also, I seem to have seen different answers for questions on advantages and disadvantages and other such tedious stuff on different papers; could this be a consequence of the mark schemes lacking the full appendices? Sometimes it seems "not random" is enough, other times "not random so biased", other times "not random so sampling errors can't be estimated"; there doesn't seem to be too much consistency...this is also lacked, it seems, in stratified sample questions; sometimes they say to number one set 1-x then the other set x+1-whatever, the other times they say to number each set starting from 1 and oe is not always present
Getting pretty worried about S3 just due to the lack of transparency with what's required; it doesn't feel very mathsy a lot of the time