Continuity correction and normal probabilities

[John H]

Can anybody help me with the following queries:

1) All questions in the Q&A and assignments appear not to use a continuity correction when carrying out a 2 sample t-test for proportions or calculating a confidence interval of the difference in proportions. The notes do not specify either way. Can anybody explain why this is and more importantly is it correct that a continuity correction is not required for a 2 sample proportion comparison?

2) When calculating normal probabilities on the calculator do we input the exact value in the calculator or simply the value to 3.d.p.

3) How many decimal places is it best to quote probabilities to after interpolation.

4) Do we need to show working for interpolation or is it sufficient to simply quote the probability from the calculator.

Many thanks in advance.

John

 

[John Lee]

Can anybody help me with the following queries:

1) All questions in the Q&A and assignments appear not to use a continuity correction when carrying out a 2 sample t-test for proportions or calculating a confidence interval of the difference in proportions. The notes do not specify either way. Can anybody explain why this is and more importantly is it correct that a continuity correction is not required for a 2 sample proportion comparison?

The continuity correction on each of the proportions would cancel out as we are doing the same inequality for both of them. So yes it is correct that it is not required.

2) When calculating normal probabilities on the calculator do we input the exact value in the calculator or simply the value to 3.d.p.

If you do 3 SF then your answer will be correct to 3 SF. I woiuld suggest using at least 5 SF (to match the Tables) then you will have all 5 figures correct.

3) How many decimal places is it best to quote probabilities to after interpolation.

If we're talking about the normal distribution again, the most you can ever quote is 5 SF as this is the limit of the accuracy provided by the Tables. Examiners solutions vary between 3 and 5 SF.

4) Do we need to show working for interpolation or is it sufficient to simply quote the probability from the calculator.

As I understand it the interpolation is the answer mark so no don't bother (unless you're worried that you're doing it wrong then you could always at least write down the 2 values and their probabilities that it's between).