The short answer: The norm should be 50% on average. When the button is blue it appears to be a negative deviation from the norm. When the button is red it is a positive variation from the norm.
"The percentages are how likely the data is to be random. There are subtle variations in color so there can be a mix of green and yellow.
Blue is anti coherence, with a percentage greater than 90%
Green is no coherence, with a percentage between 40% and 90%
yellow is a little bit of coherence, with a percentage of 10% to 40%
orange is moderate coherence, with a percentage of 5% to 10%
red is strong coherence, with a percentage of less than 5%" source:
http://gcp.djbradanderson.com/The tapestry is the visualization of the egg results for the previous five minutes.
The (very long) explanation (as I understand it):
The boxes can generate two responses 1 or 0. If you think of it like a coin toss then it would be reasonable that if you flipped a coin 100 times it would be heads 50 times and tails 50 times. Now let's have 46 coins flipping or, in this instance, boxes emitting 1 or 0. What the GCP is trying to do is look at all the eggs, what they're doing (emitting 1 or 0) and then try to correlate how the eggs are doing overall. When the variation is greater than random over a period of time the egg clor changes.
Assuming that all eggs will average 50-50 over a period of time the button is the composite of how all the eggs are acting. So, what the GCP is trying to do is measure to the deviation from the expected norm. In otherwords, they are looking at a
weighted average of the number of standard deviations the combined results deviated from chance. There are a number of ways to compute combined probability standards/variations and some can be found at:
http://pages.infinit.net/rlevesqu/Syntax/MetaAnalysis/MetaAnalysisOfPValues.txt These include:
* Notation:
- k : Nr of tests being combined
- pi: pvalue for trial i
- wi: sample size (or any weight assigned to each pvalue)
*****
* A) Fisher's Chi-square:
Chi-square(df=2k) = -2Sum
* B) Mudholkar & George's t:
t(df=5k+4)=-sqrt((15k+12)/((5k+2)k(3.14159265^2))Sum
* C) Stouffer's overall Z:
p-values are first converted to Zi
pi ---> Zi (COMPUTE zscore = IDF.NORMAL(1-pvals,0,1))
The GCP appears to be using a "Stouffer" method such that
if the null hypotheses are always true, this statistic follows a standard normal curve. Stouffer provides a measure of how many standard deviations from chance the combined results of all the units reporting to the main databse. Using a Stouffer equation (see above) a probability-value "z" can be computed which gives the
probability of observing such extreme results if
chance alone is the only explanation for the variation from the statistical standard. So, whenever the egg units exceed the normal variations, that is the the probability standard, than the composite button changes color to reflect this.
From
http://noosphere.princeton.edu/gcindex.html we get the explanation of what the color coding represents:
***
The server calculates the probability of Chisquare (the squared Stouffer Z, or Z²) on a second time frame basis (in real time) following the GCP analysis instructions. Then a Mac Widget (mini application) pulls the statistical analysis from the server and presents the user with a colored button for the probability.
***
The program creates a Chisquare for the most recent hour (the sum of 3600 seconds' Z² values)(my comment: i.e., Stouffer z squared). The probability of the Chisquare is then calculated. This value, representing the average departure of an hour of data from expectation, is recalculated each second in a sliding window, adding a new second and dropping the trailing second. Thus, the display changes very slowly, and gives an impression of the state of the network averaged over a substantial period.
The color coding represents the level of coherence or correlation among the eggs, which is reflected in the probability of the Chisquare. The expected level is about 50%, and big shifts in either direction are notable. The GCP's formal testing looks for increased interegg correlation, which is represented here by the warm colors, orange and red.
* Blue starts to fade in at 90% and above.
* Green represents about 50%
* Yellow starts fading in from green at 40%.
* Orange fades in at 15% or so.
* Red is 5% which is regarded as "significant".
* Bright red is 1%, or odds of 1 in 100.
The last sentence is the one that I have trouble understanding so I've just have been looking at the extremes as a predictor.
To continue, looking at
http://noosphere.princeton.edu/tapestry.html we learn that "The egg scores are shown as
warm color dots (reds, yellows) for positive deviations and
cool colors (blues, greens) for negative deviations."
I'm not really sure if we're supposed to be looking for positive or negative deviations, so I was reporting when it went to one extreme. The first time I mentioned the button turning blue Digit noticed that shortly thereafter it went red and within the day the earthquake occurred in Pakistan. I thought I would be around more this afternoon when I noticed the egg had turned blue and had remained blue for a while (I check on it every once in a while). Unfortunately I got a phone call that necessitated that I end up calling half my relatives to let them know another had relative passed away, so my time has been limited today. However I did notice that CNN reported a powerful aftershock had hit Pakistan when I was able to catch a breather while waiting for my dad to come to the phone. Did the egg go "red", before the aftershock hit? I honestly don't know. I look at the blue as "the quiet before the storm" and the red as the Newtonian "opposite and equal" reaction to the blue (and vice versa). I've only mentioned the egg/button being blue when it goes blue and maintains that for a while (more than a few minutes).
Anyway, does this explanation make sense? It's been a long time since I've had a mathematics course (let alone try to explain it) and I've avoided statistical analysis since I decided that I would be better off pursuing a major in history rather than an economics degree. My husband has a BSA in Physics so he's been a big help. He's much better practiced because he spent the last couple of years tutoring his oldest daughter in stats.