On the choice of a random variable which eliminates the bias caused by variable spacing in geodetic levelling

Computer values of the mean square error of the Nigerian Geodetic Levelling based on the model x_ij = ρ_ij/R _ij ^1/2 showed a uniformity between the lines of levelling, of a nature which suggest that x_ij = ρ_ij/R _ij ^1/2 is a good fit for the data. Values of the mean square error of the levelling based on w_ij = ρ_ij/R_ij showed such significant variation between the lines that this alternative model does not seem to be supported by the data used in this study.