You have a good point, Bystander, and to your point, I looked up the actual paper and they actually DID take the position in the sky into consideration in the statistical analysis which amounts to what you are suggesting.
http://arxiv.org/PS_cache/arxiv/pdf/080 ... 3247v4.pdf
we herein refer to a clockwise and anti-clockwise classification as Z and S-wise respectively
We wish to establish the large scale statistical properties of the galaxy spins. Although there is some level of uncertainty in the overall (S, Z)-wise number counts, it is still possible to look for a dipole, for example, in the spin distributions.
Rather than using an averaged map, such as that in Figure 3 we fit a probability model to all of the galaxies. The null hypothesis that we wish to test against is that we are equally likely to observe an Z or S-wise galaxies wherever you look on the sky - i.e. statistical isotropy. However, we wish to constrain alternative models in which the probability of observing a Z or S-wise galaxy varies with position on the sky ( the sum of these probabilities must always equal one).
This is not assigning a specific value of angular momentum to each galaxy but it the statistical equivalent of doing so in this situation.
Their bias and how they tested for it is also rather interesting.
They actually found a bias in favour of S spin
In this paper we are interested in the projected spin classifications of the spiral galaxies; the galaxies classified asclass=2 or class=3. We find (17,100, 18,471) of the spiralgalaxies are cleanly classified (i.e. over 80% weights) with(Z, S)-wise winding sense respectively, with equivalent (Z,S)-wise number counts of (6,106 , 7,034) for the supercleansample.
For a null hypothesis of statistical isotropy we would expect (Z, S)-wise handedness to be equally likely across the sky, however we observe a significant excess of S-wise galaxies in both our samples, at more than 7 significance2.
The original study had been of 291,626 galaxies. So they created mirror images of 93,000 galaxies from their original data and had them separately categorized by the same users. The ended up with the same S bias, and to the same degree from which they concluded that the bias was likely neurologically inherent.
We find that the average (Z, S)-wise weights are (5.5%, 6.0%) for the monochrome images. . . . The fact that the weights stay the same within statistical accuracy indicates that we have a significant level of bias in our results, and no true S-wise excess.
By the way, the original data set from which the spiral galaxies were drawn was 893,212 galaxies from 85,276 users. If I recall correctly, I read elsewhere that the Galaxy Zoo likes to have about 50 responses for each galaxy in order to eliminate errors and use results that have a high level of confidence.