Research News is delighted to announce a new regular column that will be available online each month called Stats tips. You can email your questions to editor@amsrs.com.au and each month, we will feature the best five questions and their answers. Answers are supplied courtesy of Statistics@Swinburne.

Questions should:

• be of interest to a wide audience of market or social researchers
• be of a statistical or specific methodological nature
• relate to a fairly straight forward statistics issue
• require a short response, up to one paragraph
• not require detailed calculations.

The inaugural question

Q. When you're trying to ensure the results of a study are representative of the entire Australian population, why it is generally unnecessary to interview more than 2000 people?

A. Assuming that your sample of 2000 is randomly selected, any results reported will be accurate to within +/- 2% (approximately) with 95% confidence for a population the size of Australia, or even the size of Melbourne. For example if in a survey of 2000 people we found that 55% of respondents snack between meals, then we would be 95% confident that the percentage of people for the whole of Australia who snack between meals is within the range of 55% +/- 2 % (i.e. between around 53% and 57%.)

Email your statistics question to editor@amsrs.com.au. For more information about Statistics@Swinburne, visit www.swin.edu.au/statistics.

Print this page



Other Articles in this edition