Gait analysis studies involve continuous curves of data measured over a gait cycle. Curve analysis and interpretation require adequate statistical methods. Three principal problems may be encountered in clinical practice: (i) the reliability of gait curves for a given patient, (ii) classifying a new subject as belonging to a given population or not and (iii) comparison of two populations (independent or paired). This paper presents three statistical tools for solving these problems: (i) intra-class correlation coefficients, (ii) confidence bands for a population (iii) a combination of analysis of variance and confidence bands for the difference between the means and shows how they can be used in clinical practice.