This week we have a sneak preview of a figure for one of the papers currently in the advanced stage of preparation. In that paper we are making different samples of the young stars in the HOYS survey based on their properties. This includes for example: if they are variable or not; if they show evidence of warm dust near the star or not; if they show evidence of colder dust further from the star or not; if the variation in their optical colours is consistent with extinction due to dust or not; or if they are more solar mass or below. We then develop a method to compare their variability statistics, i.e. how often do they get brighter or fainter on different timescales. This results in a probability P(A,B) that the two samples have similar variability statistics (where A, and B are the two samples).
We can use this to draw some interesting conclusions about what physical properties of the stars actually govern the optical variability. Thus, if one creates two samples that differ only in one physical property and P(A,B) is small, this means that this property has a large influence on the variability of the objects. Similarly, if P(A,B) is large when the samples differ in only one physical parameter, then this parameter has no significant influence on the statistics of the variability.
In the above plot we show some examples for P(A,B) for different combinations of samples. The x-axis in the plot (nA/NA) is a free parameter in the methodology and the plot is mostly to show that the exact choice for the parameter does not influence the values of P(A,B) and that we use 0.1 (dashed vertical line) to determine the final P(A,B) values.
What are the different sample we compare in the plot?
S(A)/S(B): These are two samples of variables that are chosen so that they represent as much difference as possible in their statistics. This is just to show that the methods works, in that if two samples have nothing in common, P(A,B)=0.0. 🙂
S(C)/S(D): This is the opposite test. We take all variable stars with evidence of warm dust near the star and randomly split this into two groups. Thus, C and D are identical and we can see that our method finds a probability of almost 90% that they have identical variability statistics.
S(2)/S(19): This is a comparison of highly variable stars with non-variable stars. Again, it shows that the method works as it only finds a 20% probability that the samples have the same variability statistics.
The other two are scientifically a bit more useful 😉
S(7)/S(8): Both are variable stars and they only differ in the masses / brightness. One contains the roughly 1M_sun sources, the other one the lower mass (0.5M_sun) objects. P(A,B) is about 0.75. Thus, there is a difference in how these samples vary, but that difference is not very large.
S(5)/S(6): These are both samples of variable stars. One has warm dust near the star, and one does not. P(A,B) is about 0.7. Again there are differences, and we can see that the presence of warm dust near the star has a greater influence on the change in the variability statistics than the mass/brightness of the star.
More of these, and what it all means, will be discussed in the paper when it is finished….stay tuned.