A Bayesian multi-planet Kepler periodogram has been developed for the analysis of precision radial velocity data ( Gregory 2005b and 2007 ) .
The periodogram employs a parallel tempering Markov chain Monte Carlo algorithm .
The HD 11964 data ( 2 ) has been re-analyzed using 1 , 2 , 3 and 4 planet models .
Assuming that all the models are equally probable a priori , the three planet model is found to be \geq 600 times more probable than the next most probable model which is a two planet model .
The most probable model exhibits three periods of 38.02 _ { -0.05 } ^ { +0.06 } , 360 _ { -4 } ^ { +4 } and 1924 _ { -43 } ^ { +44 } d , and eccentricities of 0.22 _ { -0.22 } ^ { +0.11 } , 0.63 _ { -0.17 } ^ { +0.34 } and 0.05 _ { -0.05 } ^ { +0.03 } , respectively .
Assuming the three signals ( each one consistent with a Keplerian orbit ) are caused by planets , the corresponding limits on planetary mass ( M \sin i ) and semi-major axis are ( 0.090 _ { -0.14 } ^ { +0.15 } M _ { J } , 0.253 _ { -0.009 } ^ { +0.009 } au ) , ( 0.21 _ { -0.07 } ^ { +0. % 06 } M _ { J } , 1.13 _ { -0.04 } ^ { +0.04 } au ) , ( 0.77 _ { -0.08 } ^ { +0.08 } M _ { J } , 3.46 _ { -0.13 } ^ % { +0.13 } au ) , respectively .
The small difference ( 1.3 \sigma ) between the 360 day period and one year suggests that it might be worth investigating the barycentric correction for the HD 11964 data .