Three approaches to resolving problems arising from assumption violation during statistical analysis in geographical research