The next step is to plot a correlation log k versus log P for appropriate reference substances withlog P values near the value expected for the test substance. In practice, from 6 to 10 reference substancesare injected simultaneously. The retention times are determined, preferably on a recording integratorlinked to the detection system. The corresponding logarithms of the capacity factors, log k, are plotted as afunction of log P. The regression equation is performed at regular intervals, at least once daily, so thataccount can be taken of possible changes in column performance.