Which measure of central tendency should be used for this situation?

Question

A researcher is interested in analying the increase in systolic blood pressure caused by a certain medication. Would mean, median or mode be a more appropriate measure of central tendency?

Answer 1

I would hope that something like this would be analysed a little more carefully than just a single figure. ;-)

It also depends on whether you're talking about the measures being applied to the blood pressure readings or to the increases in various patients' blood pressure. The median increase is likely to be the most useful number to generate; the median blood pressure would not be very useful, because it's not relating before and after values for a particular patient. If you're applying the measures to the blood pressure I would select the mean; if applied to the increase, the median.

However, the mean increase would contain more information about outliers which could represent significant health risks (for instance, if some groups of patients had an abnormally high increase in blood pressure). We would definitely want to bring these out in the analysis, so the mean increase might well be the most useful measure.

(In reality, we'd look for areas in the increase values where the data seemed to be clustered, and break it up into ranges accordingly, and give data for each range.)

Answer 2

Central tendency is the mean, median, and mode. It's just like another word for all three. The question is asking you, which measure would be best to use. Would you use mean, median, or mode for the movies seen? That's what's it's asking.

Answer 3

Depends:

the mean will give you the average over the entire set of measuments
the median will give you the blood pressure number that is exactly in the middle of the range
the mode will give you the blood pressure value that is encountered the most

So, depends the figure of merit you are going after.

Answer 4

No way mode, that tells you nothing. Median doesn't tell you trend. Definitely mean. It may be best to eliminate the weirdest 5% on both sides (fliers that fall waaaay outside of the norm) before taking the final mean.