It is immediately obvious from Figure9 that none of the sources we identified closely resemble Car. Their typical luminosities of correspond to stars (Maeder & Meynet1988; Maeder & Meynet1987; Meynet et al.1994; Maeder1981; Stothers & Chin1996) rather than the higher masses usually associated with LBV outbursts. Since we identify a significant population of fainter stars, this is unlikely to be a selection effect, and we conclude that these galaxies contain no analogs of Car.
There are two ways we can interpret the result. First, we can ignore the existence of Car, and set . Alternatively, we can acknowledge the existence of Car, in which case , since Car passes our selection criterion and mid-IR surveys of our Galaxy for objects as luminous as Car are probably complete. For the first case, the 90% confidence upper limit is where the period over which such systems can be detected is scaled to years. For the second case, where we include Car, we find that with at 90% confidence. In either case, the rate of transients comparable to Car is a small fraction of the supernova rate.
Stars as massive as Car are also rare, representing only to of all massive stars for a mass range from / to . If every sufficiently massive star had one eruption, the results including Car correspond to a minimum mass of ( ). If every star has an average of two eruptions, the mass limits rise to ( ). Similarly the upper limit from ignoring the existence of Car corresponds to for an average of one eruption or for an average of two. Kochanek (2011) estimated that the abundance of lower optical depth shells found at m around massive stars in the Galaxy was roughly consistent with all stars more massive than having an average of two eruptions, corresponding to , which is consistent with the present results, but close to the upper limits.