The ultimate goal of a simulation is a realization of the data given a hypothesized model of the sky and the best understanding of the detector. Since the simulation creates one of many possible realizations, re-running the simulation will result in a different realization if the generation of randomness has been altered (e.g., by using a different seed for the random number generator).
Observation simulation is part of parameter fitting, which was discussed in the likelihood section. There a physical model of the sky is folded through the detector response to calculate the expected count model. This count model is a probability distribution function that is a continuous function of the observables such as apparent energy and apparent position on the sky. In fitting this distribution function is used to calculate the likelihood value for the observed counts. Here, the distribution function is used to create a realization of the data, i.e., integral counts at specific values of the observables.
In the case of the LAT, we need to provide models of where the LAT was pointing and the gamma-ray emissions of sources within the LAT's field-of-view (FOV) over the period of the simulated observation. As was discussed in the section on LAT data products, the time history of the LAT pointing is provided by the 'spacecraft file' (also known as an FT2 file). An existing spacecraft file may suffice for your simulation since you may be interested in a simulating the data for a specific observation that has already occurred (and therefore the actual spacecraft file can be used) or for an average survey observation (in which case an spacecraft file of the correct duration can be used if Fermi was in survey mode). However, you might want design an observation, in which case you need to model Fermi's orbit and the LAT's pointing relative to this orbit; this is done with the gtorbsim tool, as discussed in the next section. The model of the sky is provided through an XML file that can be created and edited by the ModelEditor tool, which is discussed in a subsequent section. This model includes information about the source, such as: integral flux, spectral model (e.g., power law), spectral parameters (e.g., photon spectral index for a power law model), energy range, and source position (e.g., RA, DEC).
The gtobssim tool then uses the spacecraft and model files to create an event file (.FT1) that can be analyzed with the same tools that can be applied to real data. For a given LAT pointing history and sky model there is an expected gamma-ray flux incident on the LAT. gtobssim samples this expected gamma-ray flux to create a photon list. The analytic response functions are then applied to map this photon list into a observed count list: the effective area provides the probability that the photon is detected, and the apparent origin and energy of the detected photon is calculated from the point-spread function and energy redistribution function, respectively. Because the creation of the photon list and the mapping of the photon's parameters onto the detected count's observables are probabilistic processes, the code uses a random number generator, and gtobssim runs with different seeds will produce different count lists.
gtobssim uses XML model files to generate source photons with appropriate directions. These models are generally characterized by many more parameters than the actual data will justify statistically, and therefore the data will generally be fit by gtlike using simpler models. For historical reasons the gtobssim XML model files and the gtlike XML model files have different syntax.
The event file produced by gtobssim is the same filetype as the event files that will result from LAT observations, and therefore can be analyzed in the same way. Thus you can explore the data and fit models with the same tools that you will use to study real data.