I always think when I do an “all of the above” question it is too easy! This is the end of the quiz, but read on for the explanation of the last question. Or scroll down for links to some of our other quizzes.
A: comparing effect sizes between different measurement units is quite common. Comparing different units of measurement is often hard, what does 5 cm height change mean compared with a 5 kg weight change? You could, of course, report a percentile increase rather than absolute (“population X was 10% heavier than population Y but only 5% taller), but that doesn’t necessarily tell the reader the whole story; particularly if the spread of the data are different. A population that can be more spread would require a larger effect size to be meaningful. Z scores allow you and your readers to compare the effect directly.
B: comparing subjective scoring is notoriously hard. If you think about a survey response, my satisfaction score of “4” might be the equivalent of someone else’s “5” and of a third person’s “3”. If you a range of things being scored, you can use the person’s individual responses to generate a Z score for each response. This will normalise the response. Really useful for grant reviews and other situations where what matters is combining multiple reviewers into generating a ranked list of what each reviewer thinks of each thing. You could, of course, as the reviewers to rank the performances, but using Z scores allows each reviewers to have ties and to really favour or disfavour individual items, while also increases the probability that the final rank will not have ties.
C: comparing relative performance across different tests. This comes down to pretty obvious situation where test #1 has a different difficulty to test #2 and #3. Directly comparing the score of different people taking each test isn’t very useful as a discriminator. Comparing within the cohort to say someone did better or worse than average is a bit more useful but, again, the spread of the cohort means how much better is relevant. Z scores allows a direct comparison of performance relative to the cohort taking the test and their spread. Its not perfect, but saying a particular was 1SD above the mean is more meaningful than 5% above the mean.
This is the end of this quiz, I hope you found it useful. There are a load of other quizzes on this site relating to genetics and cell, molecular biology. You might also like my guidebook for researchers. It is pretty cheap and easy to read.
LaNts and Laminins 