Speaker
Description
The Fisher Information Matrix is the most widely used tool to forecast and compare the parameter estimation capabilities of future gravitational waves detectors. This method relies on a linear approximation of the detectors' response which speeds up computation of confidence intervals by millions of times with respect to stochastic sampling methods. However, neglecting higher orders terms, could lead to inaccurate and biased comparisons, especially if the detector networks have very different response morphologies. We estimate the confidence intervals for 130 signals injected in simulated future generation networks, both by Fisher Matrix and by more accurate nested sampling methods. We find that the Fisher Matrix introduces biases by overestimating the errors for networks with shorter baselines and lower SNRs, artificially exaggerating the difference between the performances of the instruments, for several parameters: the fisher matrix is biased against the triangular ET design and overall biased against ET only networks when compared to ET+CE ones. We test the assumption that the Fisher matrix should become a good and unbiased approximation for louder signals, and find that the assumption doesn't fully hold for all but 1 signal in the CoBA dataset.