Through structural equation modeling, researchers can explore relationship between several constructs. In management and behavioral sciences, researchers may be interested in whether an effect exists or not, comparing effects of different constructs on the other construct and whether mediation effect (product of path coefficients) exists or not. By setting constraint about path coefficients, researchers can test the above specific hypotheses with chi-squared difference test. The research proposed a power analysis procedure for testing constraint about path coefficients and demonstrated the procedure based on three studies in this journal. With the proposed procedure, researchers can determine required sample size to achieve desired power, and further, to reduce required sample size by appropriate research design.