Acoustic impedance inversion involves conversion of seismic traces to a reflection coefficient time series and then into acoustic impedance. The usual assumption for the transformation of poststack seismic data into impedance is that seismic traces can be modeled using the simple convolutional model. According to the convolutional model, a seismic trace is a normal-incidence record, which is an assumption that is strictly true only if the earth structure is composed of horizontal layers. In the presence of dipping layers, such an assumption is violated, which introduces bias in the result of impedance inversion. I implemented impedance inversion in the stratigraphic coordinate system, in which the vertical direction is normal to the reflectors and seismic traces represent normal-incidence seismograms. Tests on field data produced more accurate and detailed impedance results from inversion in the stratigraphic coordinate system, compared to impedance results using the conventional Cartesian coordinate system.