mescla.datasets.synthetic.carrera_application2

mescla.datasets.synthetic.carrera_application2(n_samples=100, noise='low', seed=0)

Carrera et al. (2004), Application 2: three end-members, five species.

Table 3 of the paper. End-member analyses are noisy (two levels are published, "low" and "high") while mixed samples are precise (sd = 4). This is the configuration that demonstrates the central claim: maximum likelihood improves with the number of samples, least squares does not.

Parameters:

• n_samples (int, default 100) – The paper reports results for 4, 10, 20, 50 and 100 samples.

• noise ({"low", "high"}) – Which published end-member standard deviations to use.

• seed (int, optional)

Return type:

SyntheticMixture