We present an analysis of the procedure for plasma electron number density, Ne, diagnostics based on the comparison of theoretical and experimental shape or width of hydrogen Balmer lines. Low Ne diagnostics, requiring an extension of available theoretical Stark broadening data tables was examined first. The difficulties encountered during experimental line profile analysis at low Ne are discussed and appropriate procedures suggested. The widely adopted deconvolution of experimental profile by fitting with a Voigt function is examined and it is shown that this deconvolution introduces large systematic error in Stark width determination. This uncertainty can be decreased but never completely avoided by fixing the Gaussian part of the Voigt function. Simple formulas of satisfactory accuracy for deconvolution at the half width of experimental profile were investigated and their application recommended. The contribution of Van der Waals broadening to the experimental profile is examined and the correction for its contribution discussed. Approximate and reliable formulas for the evaluation of Ne from the Stark width were critically evaluated. To estimate the applicability of different sets of theoretical data for Ne diagnostics, a comparison of theory versus experiments was carried out and best data tables were recommended. For low Ne diagnostics the application of higher members of Balmer series is advised whenever possible.