Context : The discovery of about 700 extrasolar planets , so far , has lead to the first statistics concerning extrasolar planets .
The presence of giant planets seems to depend on stellar metallicity and mass .
For example , they are more frequent around metal-rich stars , with an exponential increase in planet occurrence rates with metallicity .

Aims : We analyzed two samples of metal-poor stars ( -2.0 \leq [ Fe/H ] \leq 0.0 ) to see if giant planets are indeed rare around these objects .
Radial velocity datasets were obtained with two different spectrographs ( HARPS and HIRES ) .
Detection limits for these data , expressed in minimum planetary mass and period , are calculated .
These produce trustworthy numbers for the planet frequency .

Methods : A general Lomb Scargle ( GLS ) periodogram analysis was used together with a bootstrapping method to produce the detection limits .
Planet frequencies were calculated based on a binomial distribution function within metallicity bins .

Results : Almost all hot Jupiters and most giant planets should have been found in these data .
Hot Jupiters around metal-poor stars have a frequency lower than 1.0 \% at one sigma .
Giant planets with periods up to 1800 days , however , have a higher frequency of f _ { p } = 2.63 ^ { +2.5 } _ { -0.8 } \% .
Taking into account the different metallicities of the stars , we show that giant planets appear to be very frequent ( f _ { p } = 4.48 ^ { +4.04 } _ { -1.38 } \% ) around stars with [ Fe/H ] > -0.7 , while they are rare around stars with [ Fe/H ] \leq - 0.7 ( \leq 2.36 \% at one sigma ) .

Conclusions : Giant planet frequency is indeed a strong function of metallicity , even in the low-metallicity tail .
However , the frequencies are most likely higher than previously thought .