Grouped versus ungrouped data


Fortran must die
I run into this concept when I deal with overdispersion in GLM or logistic regression most commonly. I remain confused what the difference is between grouped and ungrouped data.

I run data that have one line per individual. Is that ungrouped data?


Fortran must die
Thanks. That reinforces my view our data is never grouped for regression, because we always analyze individual not group or aggregated data. But it makes me wonder why when you have individual data the observed variance can never be greater than that assumed in the distribution as it can with grouped data. But ultimately that does not matter I guess :p


Ambassador to the humans
Because if you only have a single observation the mean and the variance are directly related. It's impossible to have "overdispersion" - you can have lack of fit but due to the mathematics involved with only have a single observation you really can't "have too much variation" since you don't really even estimate the variance in the first place.