Methodological implications of sample size and extinction gradient on the robustness of fear conditioning across different analytic strategies