Obtaining optimal stability lobes in low immersion milling to avoid chatter requires accurate determination of the moda lparameters of the cutting system. Regression method was used previously to obtain in-process modal parameters and the result showed agreement with experiments in full immersion milling. To examine the performance of the regression method in low immersion where the width of cut is small and the effect of tooth passing frequency is more pronounced, two sets of experiments were performed with 20 and 10 mm end mill tools in three different immersions through multiple ranges of spindle speeds and depth of cuts. Different criteria were used to determine the experimental stability boundaries and it was found that the phase shift factor (𝜉𝜉) and power threshold criterion lead to more accurate chatter boundaries. Also, a novel approach was used to adopt multi-frequency algorithm instead of zero order approximation (ZOA) to perform regression method, and the estimated results showed values of natural frequency 10 to 20 Hz, and damping ratio 0.5 to 1 percent lower compared to ZOA while stiffness was generally equal or slightly higher. Besides, the predicted stability lobs by the two regression methods were compared with the experiments.