Questions people ask

§1Questions

What is this for?

Putting a galaxy relation from your own simulation next to every published simulation and observation, and finding out which of those comparisons are fair. It also lets you turn the feedback dials of two simulated universes and watch the galaxies change.

Has it been reviewed?

Not by anyone outside the project. The reviews that preceded two of the studies were internal. No outside expert has reviewed any record or result yet. A reader who knows one of these simulations could help most by reviewing some.

Can I trust the curves?

Not blindly. 55% of the simulation curves were read from published figures by scripts run by one person and have not been checked by anyone who knows the simulation. Every curve says how it was obtained, and curves read from figures are never used to rank. Where an author releases a table, use that. If you spot an error, the evidence panel has a link that opens a pre-filled report.

Why does "Check your own run" often say a curve is not comparable?

Because most published curves do not state everything needed (the IMF, the aperture, the population cut). We do not guess. If you declare your definitions and a curve declares different ones, or none, the mark says so and gives the reason. The definitions study measures how often this happens. Under the verdict, What would make more curves fair to compare? shows which answer about your run would make the most extra curves fair, so you know what to look up in your own methods.

Did you convert or correct the data (IMF, aperture, cosmology)?

No simulation curve is corrected for the IMF, the aperture, dust, which galaxies are counted or the cosmology: the differences are recorded and flagged instead. We only re-write numbers in the axis units we use, plus one h-unit conversion for the Magneticum mass functions. The observation compilation we use had already converted its curves to a common cosmology (and, for 9 of them, to a Chabrier IMF). The Explorer has an optional on-screen IMF switch, off by default; downloads never contain it. Everything is listed, with counts, in Methods §4.

Does my curve leave my computer?

No. Everything runs in your browser. The curve is stored only in your browser's local storage so it is still there when you return; the clear button removes it.

What do the dials actually do?

They scale how hard supernova winds and black hole feedback push gas out of galaxies, in two simulation codes. The curves are emulators fitted to 2,054 small simulations, so they are approximate near the edges of the dial ranges.

Can the two universes be made to agree?

On some properties, yes: you can match the stellar mass function, for example. On others, the two stay apart at every setting of the dials we tested, by more than the scatter between runs of the same code. Black hole occupation stays apart by a wide margin and gas metallicity by a small one; the best settings sit at the edge of the dials' range, where the emulators are least tested. Matching one property usually pushes others apart. The all-at-once study measures this.

Why do the two universes never agree about black holes?

Each code puts black holes in by its own rule, and none of the four dials changes that rule. The explanation page quotes the papers for both rules, shows the saved catalogues and lists what the study could not show. It explains an internal result that no outside expert has reviewed.

Is this new science?

The finding that these two codes differ structurally is consistent with what the CAMELS papers already report. What the study adds is a size for the difference in each observable's own units and a rule written before the result. See the related-work note.

How do I cite it?

Cite the dataset (DOI in the footer) and the original paper of every curve you use. The Explorer's Get data panel produces BibTeX for exactly what you plotted.

How do I add a curve or fix one?

Open an issue with the form on the repository (curve error or source suggestion). Authors who want their own curves corrected, credited differently or removed can say so there; removal is not contested.

Is it finished?

No. It is version 0.24.0 and has no outside users yet.

The full plan, with what is done and what is not, is in the roadmap.