In physics education, assessments address various forms of scientific knowledge. Most of the existing test instruments emphasise the assessment of content knowledge. These tests fail to measure the epistemic aspects of science. Thus, assessing epistemic knowledge is a theme that demands investigation. This study applies Rasch analysis to help validate an epistemic knowledge test in electricity and magnetism. Each of the initial 30 items represents distinct features of scientific reasoning. The construct dimensions are derived from Kind and Osborne's framework of Styles of Scientific Reasoning. Based on expert validation, a final 20 items were selected. The test was administered to 156 graduating in-service natural science teachers. Reliability and separation indices, person-item mapping, and principal component analysis using standardised residuals are reported. No meaningful secondary dimension was found in the test data. Invariant properties showed that the test had internal coherence. The Wright Map showed the difficulty of the test to the test takers. The overall fit statistics were within the acceptable range. Along the continuum of item difficulty, there were some notable gaps, where future test items judged to be that level of difficulty could be included. We acknowledge that there may be other reasons for these gaps.