In this post we present an update to our earlier measurement of the Baby Feeding Correlation Function:
The figure above shows a histogram of the interval between feedings for SteelyKid in the 14 weeks since birth. Error bars represent 1-σ statistical uncertainties.
As you can see, we have added a great deal of data since we last posted. The correlation function clearly shows a large peak between 3 and 4 hours, dropping off rapidly to zero for longer feeding times. The zero value for feedings within half an hour clearly demonstrates anti-bunching, indicating that baby feedings are quantized events.
We continue to await the development of ab initio theoretical calculations of the distribution with sufficient precision to compare to our data.
(The smattering of points above 7 hours is probably an artifact, either due to data entry errors, or some quirk of the script I used to sort out the data in SigmaPlot. I don't really care enough to try to figure it out. We'll just call it "background noise," and leave it at that.)
(Having figured out how to do the necessary steps in SigmaPlot, the next silly baby data project will be to re-bin the colorful feeding graphs in 23 and 25-hour "days" to see if that causes a clearer pattern to emerge. I've spent enough time puttering around with this for today, though.)
Life Science
Physical Science
Comments
points above 7 hours is probably an artifact
Definitely an artifact. The longest she's ever gone between feedings was 6.5 hours (and when I woke up and realized how long it had been, I bolted into her room in a panic to make sure she was *alive*).
Posted by: Kate Nepveu | November 18, 2008 11:44 AM
In a few days I will have my old circadian data-analysis (and graphing) software re-installed if you want me to play with the data....
Posted by: Coturnix | November 18, 2008 11:57 AM
How is the interval between the nth feeding and the (n+1)st feeding correlated with the (n+1)st feeding and the (n+2)nd feeding? (My guess is that the correlation is negative, but this is based on introspection about my own feeding habits, and I'm not a baby.)
Posted by: Michael Lugo | November 18, 2008 12:25 PM
Not that Dot Physics doesn't deserve the shot, but
Ouch!
Posted by: agm | November 18, 2008 2:55 PM
But how's the fit with an Poisson distribution?
Posted by: tcmJOE | November 18, 2008 3:56 PM
Just to let your readers know that Andor Technology plc are running a competition for the scientific community. $400 to the charity of your choice, will be awarded for the winning entry in the 2008 Christmas Card design contest, plus the opportunity to promote your work on www.andor.com - Check out:
http://www.youtube.com/watch?v=k85yzCSdhrc
Posted by: Jonny82 | November 18, 2008 5:53 PM
In a few days I will have my old circadian data-analysis (and graphing) software re-installed if you want me to play with the data....
Thanks, but about 80% of the point of this is to give me an excuse to putter around with the computer when I'm home with SteelyKid.
Not that Dot Physics doesn't deserve the shot, but
Ouch!
That wasn't intended as a shot at Dot Physics. I didn't even think of that angle until you mentioned it.
Posted by: Chad Orzel | November 18, 2008 10:37 PM
We continue to await the development of ab initio theoretical calculations of the distribution with sufficient precision to compare to our data.
Allow me to suggest an inverse Gaussian distribution, which arises as the time needed for Brownian motion to cross a fixed threshold.
Posted by: Cosma Shalizi | November 19, 2008 8:57 AM