Six weeks ago (week 242) we looked at a simple simulation of how likely it is that we can detect a periodic signal of a given amplitude in one of our light curves. We said we’d post again about this when the simulations are finished. Well, we are not there yet, but we have kept an eye on the simulation results over the last weeks. While we have now finished the simulations for only about 350 light curves, the average results are not changing significantly anymore. Thus, one can be confident that the current state of the results will very closely represent the final one.
Above is a representation of it. The right hand panel shows how the mean detection rate changes with period and amplitude when averaged over our sample of light curves. There is a clear drop of the detection rate with amplitude. This is not surprising, given that the photometric errors in our data are at best a few 0.01mag, but usually larger. There is a very subtle dependence of the detection rate with the period. For a given amplitude the detection rate drops by a few percent for longer periods. But this is a very small effect. Thus, if one finds twice as many short period objects compared to long period objects, this is real and not a selection effect in the data/detection method.
In the left panel we show the accuracy of the average detection rates we determine. One can see that everywhere the accuracy off the mean detection rates are very low – fractions of a percent. For low detection rates at low amplitudes the errors are slightly larger, but still well below one percent. This is due to the large number of repeated simulations we performed. Thus, we can trust the period detection rate values.
This will now allow us to reanalyse the period distributions and associated spot properties, as we have eliminated a potential bias/selection effect in the data. Expect to hear more about this in the coming months.