AC power lines have been considered as a convenient and low-cost medium for intra-building automation systems. In this paper, we investigate the problem of estimating the channel order and root mean squared (RMS) delay spread associated with the power lines, which are channel parameters that provide important information for determining the data transmission rate and designing appropriate equalization techniques for power lines communications (PLC). We start by showing that the key to the RMS delay spread estimation problem is the determination of the channel order, i.e., the effective duration of the channel impulse response. We next discuss various ways to estimate the impulse response length from a noise-corrupted channel estimate. In particular, four different methods, namely a signal energy estimation (SEE) technique, a generalized Akaike information criterion (GAIC) based test, a generalized likelihood ratio test (GLRT), and a modified GLRT, are derived for determining the effective length of a signal contaminated by noise. These methods are compared with one another using both simulated and experimentally measured power line data. The experimental data was collected for power line characterization in frequencies between 1 and 60 MHz.