# Averaged over different-length periods of time. Is that reliable?

In my study I will analyze how workload changes over time and over task. To do this, we record physiological measures in real-time and a subjective measure every five minutes. Moreover, we record which tasks are performed in real-time, i.e. whenever a task changes we record that.

The next step is to make a comparison. Typically, we calculate mean values across each 5-minute interval (i.e. grouped by the subjective measure) and we just assign the task that was spent the most time on in that period. The averaging of physiological values is no problem because an equal amount of samples is used.

I do not like the method of "choosing the task that was spent the most time on", however. As an alternative, I would like to average the data not by 5-minute interval, but by task-interval. The consequence of that is that some averages are calculated over a period of 1 minute, while others may be calculated over 7 minutes, for instance.

Can I assume that the averages are reliable, or are there risks to this approach (e.g. will assumptions of LMER or GEE be violated)?