How to alternate intervention - every other day? Every other week?

Pretty basic question, but I don’t think the Getting Started quite covers it.

Let’s say you want to test what an intervention (like eating kiwis or taking a pill) does to your sleep. You’ve got your lifestyle mostly constant (thanks Covid), now you just need to schedule the intervention. Blinding yourself is impossible (fake kiwis anyone?).

What are the pros and cos of,

  • test the intervention Every Other Day
  • test the intervention one week on, one week off

These two frequencies were chose to overlap with common human activity rhythms. Might there be better ones?

I’m thinking that EOD might be confused by other factors, such going to the gym every other day. On the other hand, testing the intervention for a week straight on, might lead to tolerance effects.

I’ve mostly used “always off, then months on” frequencies. Anyone suggestions?

These experiment design questions can be unexpectedly tricky to answer. Some of the issues you want to consider will require domain knowledge:

  • What do you expect are the cumulative effects?
  • What’s your guess about a washout period?
  • How difficult is the intervention to prepare and execute?
  • Are there costs of stopping? (For instance, stopping coffee can induce withdrawal effects.)

Are you literally thinking about testing Kiwis? Or was this just a placeholder?

1 Like

Definitely don’t do Every Other Day: If you are going to test something that you expect to have an immediate, non-cumulative effect (e.g. the effect of caffeine on sleep), randomize your days!

I was thinking in general, and realizing that after a number of years in QS, I don’t have a good answer to those questions, so thanks for confirming they’re non-trivial.

Specifically, some interventions I’d like to test have to do with sleep: A/C 67F vs. 75F, taking the Proper supplements, and even kiwis. Problem is that I haven’t found an accurate sleep tracker. Oura gives wildly ranging results (20 minutes vs. 60 of deep sleep) for monotonous days & nights (intermittent fasting, weightlifting in the afternoon EOD (and it’s not about the “other” day), sleep ~2am - 10am), and highly contradictory ones for outliers (e.g. great deep sleep after heavy drinking and dancing until 5am on Halloween night, followed by face planting on a random couch).

2 Likes

Probably belongs in the Oura discussion, but this is a common issue with sensor-based trackers that attempt to deliver meaningful results about complex physiological processes like sleep. Even if they function reasonable well under typical conditions, they aren’t good with atypical conditions, which are often the conditions that provoke the most curiosity. It’s such a common problem that it may point to underlying problems with how they’re conceived, what they’re supposed to be for.

2 Likes