The qpAdm Model Lab

Price

€10.00, one-time, on a published qpAdm report. One unlock covers both halves below.

Your own runs

Real qpadm() from ADMIXTOOLS 2, your sample as the target, up to 100 runs per rolling 24 hours

Sources and outgroups

Chosen by you from the same merged panel your report was computed from (AADR v66)

What comes back

The model p-value; each source's weight, standard error and z-score; the right set you chose — rejections included

The download

The exact merged EIGENSTRAT bundle (.geno/.snp/.ind) your report was computed from — multi-gigabyte, short-lived links, metered

Refined re-analysis

€15.00 per version, optional, only ever offered after publication; the original always stays

A qpAdm report answers the question the analyst asked. The Model Lab lets you ask your own: swap one source for its neighbour, add an outgroup that should break the model, try the two-way version of a three-way fit. Because the target is your own merged sample and the panel is the one your report used, every answer is comparable with the published one — same data, same method, your hypothesis.

The rejections are the point. A coordinate fit always returns percentages; qpAdm returns a p-value that can say a model is incompatible with the data. Seeing your own plausible models fail is the fastest way to understand why the published one was chosen — and, now and then, to find one worth sending back to us.

Prefer R? The same unlock includes the merged dataset itself — your kit combined with the AADR panel in standard EIGENSTRAT format — so you can run ADMIXTOOLS 2, qpWave, f-statistics or anything else on your own machine, with no run limit but your CPU. The panel version is stated, so what you compute is reproducible against what we published.

You can learn the method for free first. The AdmixTools 2 Lab runs real f-statistics, qpWave and qpAdm over the public reference panel in the browser, and Understanding qpAdm explains how to read a p-value, a standard error and a z-score. The Model Lab differs in one way that matters: the target is you.

• 01

  What exactly does the Model Lab unlock?

  Two things, for one €10.00 one-time payment on a published qpAdm report: the ability to compose and run your own qpAdm models against your own merged dataset — your sample is always the target, sources and outgroups are chosen from the same panel your report was computed from, up to 100 runs per rolling 24 hours — and the download of that merged dataset itself, the exact EIGENSTRAT bundle (.geno/.snp/.ind) your report was computed from. It is one unlock, never two purchases.

• 02

  Is it real qpAdm?

  Yes — the qpadm() function from ADMIXTOOLS 2, on our own infrastructure, over your genotypes merged with the Allen Ancient DNA Resource v66. You get the real output: the model p-value, each source's weight, standard error and z-score, and the right set you chose. Expect rejections; qpAdm refusing a model is the method working.

• 03

  Can I use a different target, or someone else's kit?

  No. The target is always the sample on the report the unlock belongs to. qpWave, f3 and f4 against a customer's kit stay operator-only; the free AdmixTools 2 Lab runs those methods over the public reference panel, never over a customer's sample.

• 04

  What is in the download, and what do I need to use it?

  Your kit merged with the AADR panel in standard EIGENSTRAT format — .geno, .snp and .ind — ready for ADMIXTOOLS 2 or any compatible toolchain on your own machine. It is a multi-gigabyte archive; download links are short-lived and downloads are metered.

• 05

  What is a refined re-analysis?

  An optional second version of your published model, only ever offered after publication. When the analyst revisits a report and finds a better model, it is published as a second version beside the original and offered as a €15.00 one-time unlock per version — never swapped in silently, never charged without your choosing it. The original stays exactly as published; once unlocked you switch between versions from the report, and every version carries its full statistics and complete right set.

• 06

  Do I need the Model Lab to trust my report?

  No. Every published qpAdm report already shows its p-value, per-source weights, standard errors and z-scores and the full right set — everything needed to re-run it yourself. The Model Lab is for going further: testing your own hypotheses on the same data, or taking the merged dataset into your own toolchain.

• 07

  Why is the search for the published model done by a person?

  Because automated model rotation has a high false-discovery rate — we built it, measured it and removed it. The published model is selected and checked by hand; the Model Lab is where you, knowing your own question, run the models you want to see.