1
$\begingroup$

I have performed a small pilot study in order to investigate whether my experimental method produces lesions of a consistent size between animals. The mean and standard deviation of lesion volume was 19 +/- 10. So, I've concluded that, with such a high standard deviation, my method did not produce consistently sized infarcts. I was wondering, how is the best way to report this, formally? Is it appropriate to say something like "With a coefficient of variation of 52%, the lesion size was inconsistent between animals". Or is it not appropriate to use CV for a single data set? Or is it more applicable to just say that, because the standard deviation is high, the lesion volume was inconsistent between animals Thank you

Kim

$\endgroup$
4
$\begingroup$

First, I'd look at the box plot and some more numbers (quantiles, outliers etc). Then I would decide:

1) Should I report any measure of central tendency at all? For example, if the distribution is bimodal or multimodal, I think that any such measure is misleading.

2) If I want a measure of central tendency, is it the arithmetic mean, the median, the trimmed mean or (possibly) some other measure? (Your example doesn't seem to call for the geometric or harmonic mean, but others might).

3) Given that measure of central tendency, what measure of spread is good? Mean usually goes well with standard deviation. Median goes well with interquartile range. There are other measures, too.

4) Regardless of any of the above, I would strongly consider showing a graph of the data. Maybe a boxplot, maybe a kernel density estimate, or, if N is small, a strip plot.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.