This paper presents a comparison of three methods to estimate the latitudinal resonance width of field line resonant, ultra-low frequency waves detected at the ground. These are a spatial domain, full-width half-maximum method and frequency domain amplitude-phase and amplitude-division methods. These methods were used to estimate the resonance width of several field line resonant intervals occurring on 26 November 2001, 1 October 2012, and 19 June 2015. The 19 June 2015 interval used data from one low, two mid, and one high latitudes. It was found that the resonance width estimates were different for each method and with how the data were processed. The most suitable methods and data processing were determined from a damped driven harmonic oscillator model. The amplitude-division methods yielded the most accurate results when the ground magnetic field data were processed with a boxcar window function or a frequency domain, exponential smoothing taper. The amplitude-phase method tended to underestimate the resonance width. The full-width half-maximum method gave accurate results for a high spatial resolution linear piece-wise curve fitted to the spectral amplitude with latitude profile. An accurate estimate of the latitudinal resonance width requires a correct choice of data processing, estimate method, resonance profile with latitude, and resonance model.