Test of the linearity of the maximum likelihood and thresholding methods: they both behave very linearly, as expected, but with different background levels. The thresholding method correction reports a higher background level and it breaks down for λ higher than approximately 1.5, whereas the MaxLik method is linear up until approximately λ = 10 (not shown on the graph). The uncertainty on λ rises with λ as expected. The Bayesian Inference method seems to slightly underestimate λ. This method contains more parameters than the thresholding method, one of them being the EM gain. A slight overestimation of the EM gain would explain the underestimation of λ.