Kimmo Sorjonen1, Gustav Nilsonne2,3, Bo Melin2, Michael Ingre2,3,4. 1. Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden. kimmo.sorjonen@ki.se. 2. Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden. 3. Department of Psychology, Stockholm University, Stockholm, Sweden. 4. Institute for Globally Distributed Open Research and Education (IGDORE), Stockholm, Sweden.
Abstract
OBJECTIVE: The present simulation study aimed to assess positive predictive value (PPV) and negative predictive value (NPV) for our newly introduced Accounting for Expected Adjusted Effect test (AEAE test) and compare it to PPV and NPV for a traditional zero-order significance test. RESULTS: The AEAE test exhibited greater PPV compared to a traditional zero-order significance test, especially with a strong true adjusted effect, low prior probability, high degree of confounding, large sample size, high reliability in the measurement of predictor X and outcome Y, and low reliability in the measurement of confounder Z. The zero-order significance test, on the other hand, exhibited higher NPV, except for some combinations of high degree of confounding and large sample size, or low reliability in the measurement of Z and high reliability in the measurement of X/Y, in which case the zero-order significance test can be completely uninformative. Taken together, the findings demonstrate desirable statistical properties for the AEAE test compared to a traditional zero-order significance test.
OBJECTIVE: The present simulation study aimed to assess positive predictive value (PPV) and negative predictive value (NPV) for our newly introduced Accounting for Expected Adjusted Effect test (AEAE test) and compare it to PPV and NPV for a traditional zero-order significance test. RESULTS: The AEAE test exhibited greater PPV compared to a traditional zero-order significance test, especially with a strong true adjusted effect, low prior probability, high degree of confounding, large sample size, high reliability in the measurement of predictor X and outcome Y, and low reliability in the measurement of confounder Z. The zero-order significance test, on the other hand, exhibited higher NPV, except for some combinations of high degree of confounding and large sample size, or low reliability in the measurement of Z and high reliability in the measurement of X/Y, in which case the zero-order significance test can be completely uninformative. Taken together, the findings demonstrate desirable statistical properties for the AEAE test compared to a traditional zero-order significance test.
Authors: Brian M D'Onofrio; Arvid Sjölander; Benjamin B Lahey; Paul Lichtenstein; A Sara Öberg Journal: Annu Rev Clin Psychol Date: 2020-05-07 Impact factor: 18.561