Following the old adage that the fish rots from the head first I decided this evening to check out some more of the paper by the Kroupa group, particularly Kroupa’s recent first author contributions. And indeed I quickly found another example of dubious astronomy based on the careless use of third-party datasets with this paper (Kroupa & Petr-Gotzens 2011).
The authors’ aim is to reconstruct the Initial Period Function (IPF) of binaries amongst the population of young stars and to thereby search for variation in shape (and normalization) relative to that of old stars, with the ultimate goal being to motivate theoretical models for the evolution/disruption of these systems. A key plot from this paper is their Figure 1 in which the authors present a reconstruction of the IPF in Taurus-Auriga compiled from the third-party datasets of five separate observational studies sensitive to mostly quite different ranges of period (spanning a few orders of magnitude). The problem is (much like in the Pflamm-Altenburg et al debacle) that very little care is given to the handling of these datasets.
For instance, the samples of Leinert et al and Kohler et al are truncated to lPs (log10 periods [in days]) between 4.5 and 7.5 and combined directly to estimate the IPF over this interval despite the fact that inspection of each dataset separately suggests that perhaps one or both are not uniformly sensitive over this range. As I illustrate through the “normalized” (to a maximum bin height of 1) histograms of lPs in these datasets shown below, the Kohler et al sample strongly favours long periods whereas that of Leinert et al favours intermediate periods. And indeed a simple KS test advocates rejection of the null hypothesis (that these represent samples from a common parent distribution) at a p-value of 0.03. Such non-uniform sensitivity (or “selection bias”) could well be present in the Richichi et al sample (from which the authors only take the 3 < lP < 4 period detections), but the point is that we really have no way of telling from the available data.
Seems fishy to me …