Hello Jacques,

It's an interesting question, but I wasn't completely clear as to whether you were interested in comparing the means of within-person meal size variation, or comparing the variances of mean person meal size across groups.

If the former, you could compute within-person meal size variation via a standard deviation (SD across all daily meal averages or totals for the number of days you monitor), or an interquartile or semi-interquartile range (if you thought that these values wouldn't necessarily be of interval strength). Then compare the means of the spread statistics via a t-test, or possibly ancova (body mass might be a useful covariate; sex might be worth building in as a blocking factor; age possibly could be a covariate as well).

If the latter, then each person's datum is the average daily food intake (however you quantify this; perhaps as kiloCalories), then a two-group variance comparison, such as Hartley's Fmax, Levene's test, or some other procedure could be used.

There would be fancier ways to proceed, especially if you had some intervention going on through the course of the observations for one or both groups.

Good luck with your work.

Thanks for your prompt and thoughtful response. I'm interested in the second- is the individual variability in food intake from day-to-day (sequence is irrelevant) different between two populations. So I have two groups, and a mean and variation for each of several individuals from each group. I am not interested in comparing the variance of population means but comparing the variability of individuals (a population of variance values- which I thought could not be normalized.

See these papers

Article Comparing Between- and Within-Group Variances in a Two-Level...

Article Powerful Statistical Inference for Nested Data Using Suffici...