Two questions: confidence intervals and live syncing

171 views
Skip to first unread message

Roman Duda

unread,
Nov 3, 2013, 6:48:25 PM11/3/13
to tag...@googlegroups.com
Hi all,

Just started using the android app on Thursday and I'm very much liking it!

I have two questions (I hope there isn't already an answer to these floating around, I apologise if there is!)

(1) Is there a way to calculate confidence intervals for tags? cntpings.pl only gives a best guess, but doesn't give an interval, even when it's verbose.
Is there an option within cntpings.pl to get confidence intervals?
Is there another script in TagTime for getting confidence intervals?
If the answer to both of the above is no, does someone have at hand an idea of how to calculate a confidence interval from tagtime data? (I have some experience with R, so a pointer as to how to do it in R would be ideal.)

(2) Is there a way (I'm guessing no, but though it was worth checking) to automatically get the log to sync with a file on my computer? At the moment I upload the log to google drive, and then run the cntpings script on it. If there was a way of not having to do this, that'd be awesome, or at least to be able to be pushed to the same filename, that'd also be convenient.

Thanks so much for putting this together!

Roman
 
 

Nicolas Felix Hogan

unread,
Nov 9, 2013, 7:56:06 AM11/9/13
to tag...@googlegroups.com
Makes me wonder if you could give TagTime permission to modify a Google Doc so that you could have it automatically update every time it synced. Would be an amazing feature.

An option to view the analysis of the log online, a website version of cntping if you will would be the bee's knees.

Daniel Reeves

unread,
Nov 13, 2013, 3:54:41 AM11/13/13
to tag...@googlegroups.com
I have Mathematica code for #1:

a = 4/3; (* TagTime's gap is 3/4 hours and the rate parameter is 4/3. *)

(* The CDF for the sum of n exponential random variables with rate
parameter a. Eg, the probability that 6 pings will happen within 8
hours is cdf[4/3,6,8]. *)
cdf[a_, n_, x_] := 1 - GammaRegularized[n, a*x]

(* What x makes cdf[a,n,x] equal to p? *)
cdfi[a_, n_, p_] := InverseGammaRegularized[n, 1 - p]/a

(* Gives c-confidence interval on the sum of n exponential RVs. *)
ci[c_, a_, n_] := {cdfi[a, n, (1 - c)/2], cdfi[a, n, c + (1 - c)/2]}

So, for example, the 95% confidence interval on the amount of actual
time spent getting 10 pings is
ci[.95, a, 10] = {3.6,12.8} hours

For #2, you can now have certain sets of pings auto-update Beeminder,
which is not exactly the same as having the whole tagtime log file
automatically synced somewhere. Working on that!
> --
> You received this message because you are subscribed to the Google Groups
> "tagtime" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to tagtime+u...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.



--
http://dreev.es -- search://"Daniel Reeves"
Goal tracking + Commitment contracts == http://beeminder.com

Roman Duda

unread,
Nov 26, 2013, 7:49:00 PM11/26/13
to tag...@googlegroups.com, dre...@beeminder.com
Sweet, thanks Daniel! I'll try to recreate this in R! :)
Reply all
Reply to author
Forward
0 new messages