Abstract
The primary results of most observations of cosmic microwave background (CMB) anisotropy are estimates of the angular power spectrum averaged through some broad band, called band-powers. These estimates are in turn what are used to produce constraints on cosmological parameters due to all CMB observations. Essential to this estimation of cosmological parameters is the calculation of the expected band-power for a given experiment, given a theoretical power spectrum. Here we derive the "band power" window function which should be used for this calculation, and point out that it is not equivalent to the window function used to calculate the variance. This important distinction has been absent from much of the literature: the variance window function is often used as the band-power window function. We discuss the validity of this assumed equivalence, the role of window functions for experiments that constrain the power in {\it multiple} bands, and summarize a prescription for reporting experimental results. The analysis methods detailed here are applied in a companion paper to three years of data from the Medium Scale Anisotropy Measurement.