Amount of Data Needed

This is one of the more common questions we get and I’m sorry to say there really isn’t one answer that applies here, as one can imagine. I can point to some of our work using OpenDBM and you can look at the sample sizes in there (which are fairly small). But here are some truths. More data is of course always better; no denying that. Video and audio data, though, is quite rich, and the length of behavior one would need to see effects (in our experience) is not hours, but minutes and sometimes just seconds. Importantly, this differs depending on the patient population and the behaviors during which the measures are being quantified. The only real advice we can provide in this matter is to search the literature and see what it says. In most cases, there is precedent to lean on. That can help with selecting the number of patients, time points, and length of video/audio needed to see effects.

Good luck!