The conditions under which X squared measures the discrepancy between observation and hypothesis