Libre Time in Target calculation (Geek alert)

[Rob Oldfield]

Hi all,

As indicated above, this is an extremely geeky query about how Libres calculate the Time in Target percentage. Apologies to anyone who falls asleep reading it.

Question really is how does it handle the time period when calculating the percentage over the last 7 days. Yesterday (Wednesday) at 11am I was happy because my figure was showing as 90% and, checking back I could see that the previous Wednesday had been very iffy and I'd only been in target for about an hour from 11am to 6pm. So, I thought, if I keep in range yesterday then all that nasty out of range from the week before would be dropping out of scope. Six hours out of range should be 6/(7*24)=3.5% of a 7 day period and I'd be up to 93 or 94%. A new record high!

Over the day my levels stayed resolutely in range but did the Time in Target percentage go up? I know you'll be as disappointed as me to find out that it did not!

I'm exaggerating of course, it doesn't really matter. But I'd be interested if anyone has any more info on how that calculation and time period are handled.

All the best

 

[Bruce Stephens]

I honestly don't know. Maybe the current day doesn't get included until it's finished (so you're still seeing last Wednesday) or something? I'd expect it to have dropped off by tomorrow anyway.

 

[Rob Oldfield]

I honestly don't know. Maybe the current day doesn't get included until it's finished (so you're still seeing last Wednesday) or something? I'd expect it to have dropped off by tomorrow anyway.

Yes, I thought it might do something like that and recalculate each day or every six hours but I have seen the percentage dropping percent by percent when I've had a hyper issue been too high for a period of time.

 

[SB2015]

I don’t know but I suspect that it will be a rolling average, using the previous 7x24 hours for the averages, taken from the current time. As you say it does not matter but now that you have asked the questions .... if you get a definite answer I shall be interested to see what it is.

 

[Deleted member 27171]

I also don’t know the answer so just here to add to speculation. Did you have any data gaps? Perhaps if there are any the system doesn’t include the time as in range.

 

[Bruce Stephens]

I don’t know but I suspect that it will be a rolling average, using the previous 7x24 hours for the averages, taken from the current time.

I'm not sure exactly what you're suggesting there, but whatever it is I think I'm skeptical. 7x24 hours is only 672 readings and 90*24 hours is only 8640 (since there are 4 an hour) so I think it's very likely to be recalculated when you look at it. (The charts change plausibly when I change the range, which again would be consistent with the app and reader just recalculating.) (I'm aware that doesn't give an explanation of the original question.)

 

[Rob Oldfield]

I also don’t know the answer so just here to add to speculation. Did you have any data gaps? Perhaps if there are any the system doesn’t include the time as in range.

I did think that might be possible but not sure I believe it. Sensor usage is showing 100% data captured over last 30 days, though the last sensor change I did was on Monday so that would have been included in the seven day period. But then they'd have to have set the system up to either classify 'no active sensor' as either 'high' or 'low' which would clearly be a bit mad. Percentages always total to 100 (as far as I've seen).

 

[Rob Oldfield]

I'm not sure exactly what you're suggesting there, but whatever it is I think I'm skeptical. 7x24 hours is only 672 readings and 90*24 hours is only 8640 (since there are 4 an hour) so I think it's very likely to be recalculated when you look at it. (The charts change plausibly when I change the range, which again would be consistent with the app and reader just recalculating.) (I'm aware that doesn't give an explanation of the original question.)

Ah. Interesting that it recalculates on the fly if you change the range. I'd not tried that. Does seem to indicate that it _should_ be continually updating.

 

[mikeyB]

You can change the time in target on the reader to have a look at Time in Target for up to the last 30 days. Simply use the left arrow button at the bottom of the screen, so clearly adds in data from the previous sensor. And it will change all the figures if you change the range.

Don’t know what the problem under discussion is, to be honest.

 

[Bruce Stephens]

Don’t know what the problem under discussion is, to be honest.

If the past 7 days shows a slightly disappointing Time in Range value, but the main failure was in the first day of that, then one would expect the Time in Range to be significantly better tomorrow (since that first day would drop off the 7 days) but apparently that didn't happen.

 

[Rob Oldfield]

If the past 7 days shows a slightly disappointing Time in Range value, but the main failure was in the first day of that, then one would expect the Time in Range to be significantly better tomorrow (since that first day would drop off the 7 days) but apparently that didn't happen.

Nicely summarised. I've now seen that there's a download option in LibreView that lets you get at the raw data. I'm going to try some number crunching......