Date of Award


Degree Type


Degree Name

Master of Science (MS)


Mathematics and Statistics

First Advisor

Chris Corcoran


Family-based study designs are often employed when investigating the genetic causes of complex disease. While the transmission disequilibrium test (TDT) and its extensions were developed to use family data for assessing linkage between a known genetic marker and a disease-causing gene, the so-called FBAT approach proposed by Rabinowitz and Laird (2000) effectively subsumes these family-based procedures as special cases. FBAT is fully conditional, but its implementation in the freely available FBAT software package uses a large-sample distributional approximation to compute p-values. The exact distribution for FBAT can be enumerated, but doing so explicitly is computationally intensive, particularly for relatively larger sample sizes. Schneiter et al. (2005) proposed an efficient algorithm for computing the exact p-value, but the computational performance of this procedure has not been systematically evaluated. In this report, we carry out a simulation study to determine the relative computational efficiency of large-sample FBAT and exact FBAT (XFBAT). Characteristic of all exact significance tests in statistical practice, XFBAT is significantly more computationally burdensome, and its efficiency decreases exponentially for larger sample sizes. In addition, XFBAT is more computationally intensive under the additive inheritance model, as opposed to either the dominant or recessive models.

Included in

Mathematics Commons