A method of calibrating a gas sensor that generates an output related to the concentration of a sensed gas, the method comprising: i) obtaining the value of said output or a related value at each of a set of known values of a first variable and at each of a set of known values of a second variable, one of the variables being gas concentration; ii) for each value of said second variable, determining a first best fit function relating said output or related value to the first variable, each best fit function utilizing at least one coefficient whose value is determined; iii) for the, or the corresponding, coefficient values from each of the first best fit functions, determining a second best fit function relating said coefficient to the second variable, and replacing the coefficient values in the first best fit functions by the second best fit function, and, iv) repeating step iii) for each further coefficient, if any, in the first best fit functions so as to generate a final function relating said output or related value and said first and second variables.