The present invention relates to method for interpreting data obtained by measuring a length of optical fiber using an optical time domain reflectometer (OTDR), and comparing that measurement to a reference measurement. The technique uses statistical inference to determine a likely cause for a length measurement to be shorter than a reference length. One technique uses a chi-squared best fit of an array reflectance spike occurrences along the fiber to a historical reference array. In that way, it can be determined whether the missing portion of the tested fiber is at an end or between the ends, providing evidence that the short length measurement results from a fiber break or from the intentional removal of a reserve loop.