Strategy for method validation: assumption of linearity and simple use of the Hartley's test to decide the type - weighted or unweighted - of the linear regression. Example of the plasmatic quantitation of midazolam by HPLC.We have described a complete and simple procedure for linear regression analysis used in bioanalytics. Firstly, the equation of the least squares line was calculated. Linearity was then carefully investigated using a residuals's study and analysis of variance which test the quality of the linear model. After verification of the linearity, the homogeneity of the variances is tested by the mean of the Hartley's test. From this information, the type of the calibration line - weighted or non weighted - was decided. Computation of weighted regression line was explained. This weighted model is requested in case of non homogeneity of the variances, i.e. when the range of calibration is wide, especially for pharmacokinetic studies as it is the only model that gives correct estimation of the standard deviations and confidence limits. Accuracy was evaluated by the mean of a t-statistic. Limits of decision, detection and quantification were calculated from an analysis of the blanks. This procedure was applied for illustration to the quantitation of midazolam in plasma by high performance liquid chromatography. Midazolam was extracted from 1 ml of plasma by solid phase extraction on C8 cartridges and recovered by 0.5% acetic acid in methanol. Chromatography of the extract was performed on a reversed phase C18 column. Analysis was achieved within 10 minutes.