Testing the Roman era sample from Marathon, Greece with qpAdm, f4 statistics and G25

For people wondering why Marathon on G25 appears more Levantine on Neolithic models:

1) The G25 sample has high distances because something is wrong with the cords.

2) proximal G25 model shows it identical to Roman Mugla, G25 is better at proximal modelling.

3) F4 show that Levant Sidon does not have more affinities to Marathon than Mugla.

4) qpAdm shows that it cannot be modelled with Mugla + Levantine but it can only be modelled as 100% Roman Mugla.

Conclusion: The Roman Marathon sample is almost identical or even identical to Roman Mugla:

qpAdm and F4:

right = c('Mbuti.DG', 'Russia_Samara_EBA_Yamnaya', 'Russia_Karelia_HG', 'Serbia_IronGates_Mesolithic', 'Turkey_N', 'Iran_GanjDareh_N', 'Spain_IA', 'Greece_Minoan_Lassithi', 'Croatia_MLBA', 'Netherlands_EIA', 'Netherlands_MBA', 'Russia_IA_Ingria.SG', 'Latvia_BA', 'Israel_C', 'Armenia_EBA_KuraAraxes', 'Mongolia_North_N', 'Russia_MLBA_Sintashta', 'Iran_C_SehGabi', 'Turkey_EBA_II.SG', 'Turkey_EBA_Isparta', 'Turkey_OldHittitePeriod.SG', 'TUR_C_Gordion_Anc_Iron_Age', 'Croatia_EIA', 'Albania_BA_IA', 'Macedonia_Classical_Hellenistic', 'Bulgaria_EIA', 'Greece_BA_Mycenaean', 'Turkey_Archaic_SubGeometric', 'Turkey_Alalakh_MLBA', 'Lebanon_MBA.SG')

G25: