The Figure from Kubinger (2005) plots the Rasch item calibrations for 481 males against those for 439 females. Overall, the differences are statistically significant. But Kubinger remarks:
"However, the differences in parameter estimation are of no practical relevance. The differences are too small to impose any systematic handicap on an examinee's test result because the standard error of estimation of the ability parameter is much greater than errors due to the given item parameter differences. Goethals (1994) provided a rule of thumb: Any difference of parameter estimations should not exceed a tenth of the range of the parameters."
Klaus D. Kubinger (2005) Psychological Test Calibration Using the Rasch Model - Some Critical Suggestions on Traditional Approaches. International Journal of Testing, 5:4, 377 - 394
Goethals, R. (1994). Die praktische Erprobung von Alternativen zur multiple-choice-Vorgabe bei Computertests [Experiences with alternatives for traditional multiple-choice response formats in computerized testing]. Unpublished doctoral dissertation, University of Vienna, Austria.
Goethals' Guideline for Item Calibration Invariance … R. Goethals, Rasch Measurement Transactions, 2008, 22:2 p. 1166
The URL of this page is www.rasch.org/rmt/rmt222d.htm
Website: www.rasch.org/rmt/contents.htm