Logging Unlogged Information

Does anybody have any idea on what to do with information that you forgot to log?

The other night, I forgot to record my sleeping time, so I didn’t even bother to record the time I woke up. Now there’s a gap in the data. So, when I will write code to process and analyze the data, I must be mindful with.

1 Like

I think that there are two sides to this.

  1. Make the capture process as simple as possible so that it is not a chore to actually do the logging. For sleeping I use a FitBit Ultra, one click when going to sleep, one when I wake up; but even so, on occasion I still forget!

  2. Measure what you are measuring. On my personal dashboard, as well as logging my hours of sleep I also log the fact I have logged my sleep so that I can measure how often I forget and look for trends/causes.

Hope that helps.


I do not have the money to buy a fitbit ultra.

What are you using just now?
Pencil and paper by your bed and then a Google Spreadsheet would give you the same outcome!

A libreoffice spreadsheet.

I hope it won’t sound too technical, but you have two options:

  • data imputation: compute the arithmetic average for that variable, i.e. average sleep time from previous days, and use that number to fill in the gap; you can also use median instead of mean. For instance, you know that you slept on average 6.5 hours that week. Use that number to fill in the gap.
  • data ascription: if you have other variables recorded on that day, go through other records and find a day that was very similar to the one when u forgot to log X, and use that day’s value of X. For instance, you forgot to log your sleep time, but all the other information is there: number of hours worked out at the gym, number of steps taken, calories burned, productivity, stress and mood levels, etc. Provide that have enough records (like 2-3 months), find a day that has similar number for all these variables, and use the sleep time from that day.

I hope this helps!


I think that there is room for EEG assessed sleep, accelerometer assessed sleep, PIR assess sleep, audio assessed sleep, sleep/wake recordings and even gusestimation. There just needs to be a common way to express the fidelity of that description of the event. Writing the time you turned out the lights and the time you rise from bed the next morning provides a particular level of fidelity on a spectrum. It’s prone to a certain amount of error. Even a guess has a different spot on the spectrum. Did you guess immediately on wake? Not until Lunch? Last week? Were you intoxicated? How good is your temporal reasoning? Can you use other data to support your estimate?

It might not be valuable to scale the accuracy of your pure guesses, but the different technical methods should certainly have a standard assessment.

I think I’ll write more on this subject…

I finally wrote a post on what you can do with unlogged information: http://measuredme.com/2013/01/personal-analytics-101-how-to-deal-with-holes-in-your-self-tracking-data/