Computational modeling is used to systematically examine many of the sources of statistical variance in particle parameters during thermal plasma spraying. Using the computer program LAVA, a steady-state plasma jet typical of a commercial torch at normal operating conditions, is first developed. Then, assuming a single particle composition (ZrO2) and injection location, real world complexity (e.g., turbulent dispersion, particle size and density, injection velocity and direction, etc.) is introduced "one phenomenon at a time" to distinguish and characterize its effect and enable comparisons of separate effects. A final calculation then considers all phenomena simultaneously, to enable further comparisons. Investigating each phenomenon separately provides valuable insight into particle behavior. For the typical plasma jet and injection conditions considered, particle dispersion in the injection direction is most significantly affected by (in order of decreasing importance): particle size distribution, injection velocity distribution, turbulence, and injection direction distribution or particle density distribution. Only the distribution of injection directions and turbulence affect dispersion normal to the injection direction, and are of similar magnitude in this study. With regards to particle velocity and temperature, particle size is clearly the dominant effect.