We present an analysis of three years of precision radial velocity measurements of 160 metal-poor stars observed with Keck/HIRES. We report on variability and long-term velocity trends for each star in our sample. We identify several long-term, low-amplitude radial-velocity variables worthy of follow-up with direct imaging techniques. We place lower limits on the detectable companion mass as a function of orbital period. None of the stars in our sample exhibits radial-velocity variations compatible with the presence of Jovian planets with periods shorter than the survey duration (3 yr). The resulting average frequency of gas giants orbiting metal-poor dwarfs with −2.0≤[Fe/H]≤ −0.6 is fp < 0.67%. By combining our dataset with the Fischer & Valenti (2005) uniform sample, we confirm that the likelihood of a star to harbor a planet more massive than Jupiter within 2 AU is a steeply rising function of the host's metallicity. However, the data for stars with −1.0≤[Fe/H]≤ 0.0 are compatible, in a statistical sense, with a constant occurrence rate fp≃1%. Our results usefully inform theoretical studies of the process of giant planet formation across two orders of magnitude in metallicity.