Summary form only given. A series of gas-cell experiments were performed at NRL to be able to validate dry air chemistry/collision models available in Particle-in-Cell (PIC) codes for cavity System Generated Electromagnetic Pulse (SGEMP). The experimental setup is a co-axial power feed which transitions to a disk-shaped cathode. The electron beam then passes through a thin anode foil and then a pressure foil, which separates the vacuum diode from the gas cell; scattering and propagation in the foils is energy and angle dependent. The peak electron energy entering the gas cell is about 90 keV and the pulse-duration is about 100ns. In addition to voltage and current measurements at various locations, a feature of these experiments is laser interferometry, which measures the line-integrated electron density at various locations in the gas cell for a gas pressures ranging from 50 mTorr to 300 Torr. This paper will analyze the numerical error as well as aleatory and epistemic uncertainty. PIC and Monte Carlo electron transport algorithms are inherently stochastic in nature; therefore, an extension to Richardson extrapolation is developed to obtain an error bound on the simulations. To account for aleatory and epistemic uncertainty we use second order probability, an inner aleatory loop sampling the approximate distribution of variation in parameters in the experiment with an outer epistemic loop varying the parameters of the distribution and model form. By including both aleatory and epistemic uncertainty and numerical error, the simulation can be compared with the NRL gas cell experiment.