Why Does My Normalized Star Flux Data Show a Bimodal Distribution?

Click For Summary
The discussion centers on a researcher encountering a bimodal distribution in normalized star flux data from Tycho photometry, despite expecting a normal distribution. The researcher has converted V-band magnitudes to flux, calculated means and standard deviations for multiple photometric measurements, and normalized the data using a standard formula. They express concern that the bimodal result might be due to issues with fainter stars or other underlying astronomical factors. Suggestions include conducting normality tests to assess the distribution and consulting a statistician for further insights. Overall, the researcher seeks advice on potential causes and solutions for the unexpected data distribution.
big man
Messages
241
Reaction score
0

Homework Statement


I'm absolutely desperate for help on this. The image supplied below is an image of a frequency histogram for the V-band Tycho photometry data (from the Tycho Epoch Photometry Annex). What I've done is this:

1. I've converted the V-band magnitudes into flux so that it is no longer on a log scale.

In the Hipparcos mission there were a number of transits for each star (70-200). So that means I have 70-200 photometric measurements for each star.

2. I've now calculated the mean of all the flux values for those transits of a particular star as well as the standard deviation. This process is repeated for each star.

3. I have then normalised the data by using the following formula:

Z=(X-µ)/σ

where X is the original flux value, µ is the mean of the star’s dataset and σ is the standard deviation of the star’s dataset.

4. The resultant 'normalised' data has then plotted using Matlab (seen below).

The problem is that I'm expecting something that is at least close to a normal distribution. But instead I have this strange bi-modal function that is more narrow than a typical normal distribution.

What could cause this bi-modal effect? I have no idea at all. Well the only thought I've had is that maybe this approach fails for fainter stars. I was also thinking of making an appointment with a statistician at my university, but then I'm not sure if he would necessarily know (if it happens to be some astronomical issue).

I thought I'd try here first because there are some people here who do astronomy and might know something that I'm missing.

Any advice on this would be extremely appreciated.
Thanks








http://img521.imageshack.us/img521/4174/untitledqg6.jpg
 
Last edited by a moderator:
Physics news on Phys.org
The distribution appears to be bimodal, but appearances can be deceiving. There are several tests for normality. Have you tried any of these to determine whether you must reject the hypothesis that the distribution is normal?

Make the appointment with the statistician.