Read on Twitter
This is probably a bit weird but, I want to plug an idea from a 5-year-old paper ( https://tinyurl.com/5goydyb9 ) that has come up a lot recently in our lab. And it's simply this: you need to pay attention to the temporal structure and scaling of your measures *and* your analyses. 1/
The example pictured here illustrates the different conclusions you would draw from the MSSD statistic if you sampled once, versus 4 times per day. If you're measuring a sinusoidal process, once per day will reflect a stable system, but 4x/day will reflect an unstable system. 2/
Did you sample 4x/day? 1x/day? Was this based on theory? Empirical observation? What's the frequency of the DV? On what interval will you run the model? I worry that we "plug & chug" with time-dependent analyses like ARs & MSSDs without taking these factors into consideration. 3/
Anyway. I have lots of thoughts on the matter. For instance, in the paper linked above (wink, wink, nudge, nudge). I'd love to hear what other people think. Are you worried about this? Do you think about it when you plan a study or an analysis? 4/4
