You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+4-1Lines changed: 4 additions & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -295,6 +295,9 @@ To get the closest behavior to that of FAST-IDL, you should set ```C_INTERVAL=68
295
295
*```SAVE_SIM```: possible values are ```0``` or ```1```. The default is ```0```. If set to ```1```, the program will save the best fitting parameters in the Monte Carlo simulations of each galaxy in a FITS table located at ```best_fits/[catalog]_[source].sim.fits```. This table contains the values of all fitting parameters and the chi2. This will consume some more disk space, but will not slow down the program significantly. It can be useful to identify covariances and degeneracies that are not apparent from the confidence intervals printed in the output catalog.
296
296
*```BEST_FROM_SIM```: possible values are ```0``` or ```1```. The default is ```0```, and the best fitting solution will be chosen as the one providing the smallest chi2 value in the grid. If set to ```1```, the program will instead determine the best solution from the median of all the Monte Carlo simulations. This will ensure that the "best fit" values are more consistent with the confidence intervals (i.e., usually more centered), and erases large fluctuations when multiple solutions with very different fit parameters lead to very close chi2 values. This typically happens for galaxies with poor photometry: there are a large number of models which give similarly good chi2, but one of them has a chi2 better by a very small amount (say 0.001) and it thus picked as the "best fit".
297
297
298
+
## Confidence intervals from chi2 grid
299
+
* ```INTERVAL_FROM_CHI2```: possible values are ```0``` or ```1```. The default is ```0```, in which case confidence intervals (error bars) are computed using a series of Monte Carlo simulations of the data. An alternative, much faster way to obtain similar results is to read off the range of parameter values covered by models in the grid, selecting only the models that have a chi2 within a certain threshold from the best value (see Avni 1976). This can be done by setting this value to ```1```. While in most cases the resulting confidence intervals will be almost the same as with the MC simulations, there are several advantages to using this approach. The first is that, by construction, the resulting confidence intervals will always include the best fit value; this is not guaranteed with MC simulation. The second is that this method behaves better when the best fit solution is close to the edge of the grid; intervals from MC simulations are computed from percentiles and therefore may under-represent edge values. The latter is particularly important for parameters that have a natural upper/lower bound, such as parameters that cannot be negative but for which zero is a perfectly fine value (e.g., Av, or some SFH parameters). There are some drawbacks however. To obtain precise confidence intervals, the grid must be relatively fine: if the true confidence interval of a parameter is 1 +/- 0.01, but the fitting grid for that parameter only has a step of 0.1, then the code will report 1 +/- 0, which is too optimistic. Finally, in terms of performance, for this option to work the program has to write part of the chi2 grid on the disk, and the amount of space required can be large if you have a very fine grid.
300
+
298
301
## Controlling the cache
299
302
*```NO_CACHE```: possible values are ```0``` or ```1```. The default is ```0```, and the program will read and/or create a cache file, storing the pre-computed model fluxes for reuse. If you are changing your grid often or if the grid is very large and you do not want to store it on the disk, you can set this value to ```1``` and the program will neither read from nor write to the cache. Because it avoids some IO operations, it may make the program faster when the grid has to be rebuilt.
300
303
@@ -320,7 +323,7 @@ To get the closest behavior to that of FAST-IDL, you should set ```C_INTERVAL=68
320
323
*```BEST_SFHS```: possible values are ```0``` or ```1```. The default is ```0```. If set to ```1```, the program will output the best fit star formation history (SFH) to a file, in the ```best_fits``` directory (as for the best fit SEDs). If Monte Carlo simulations are enabled, the program will also output confidence intervals on the SFH for each time step, as well as the median SFH among all Monte Carlo simulations. This median may not correspond to any analytical form allowed by your chosen SFH model.
321
324
*```SFH_OUTPUT_STEP```: possible values are any strictly positive number, which defines the size of a time step in the output SFH (in Myr). The default is ```10``` Myr.
322
325
*```SFH_OUTPUT```: possible values are ```'sfr'``` or ```'mass'```. The default is ```'sfr'```, and the program outputs as "SFH" the evolution of the instantaneous SFR of each galaxy with time. If set to ```'mass'```, the program will output instead the evolution of the stellar mass with time (which is usually better behaved, see Glazebrook et al. 2017). Note that the evolution of the mass accounts for mass loss, so the mass slowly _decreases_ with time after a galaxy has quenched.
323
-
* ```SAVE_BESTCHI```: FAST++ can save the entire chi2 grid on the disk with the ```SAVE_CHI_GRID``` option. However, if you have *huge* grids, this can require too much disk space (I have been in situations where the chi2 grid would be as large as several TB!). Usually, one is not interested in the chi2 of *all* models, but only those that match the data within some tolerance threshold. This option allows you to only save on the disk the models that are worst than the best chi2 by some amount ```chi2 - best_chi2 < SAVE_BESTCHI``` (where ```SAVE_BESTCHI=1``` if you are interested in standard 68% confidence intervals, or ```2.71``` for 90% confidence, etc., see Avni 1976). These "good" models are saved in a separate ".grid" file for each galaxy of the input catalog, inside the ```best_chi2``` folder. The format is similar to the ".grid" file for the full chi2 grid (which is described above), but not identical. The ``fast++-grid2fits`` tool can also convert these files into FITS tables. The binary format is the following:
326
+
* ```SAVE_BESTCHI```: FAST++ can save the entire chi2 grid on the disk with the ```SAVE_CHI_GRID``` option. However, if you have *huge* grids, this can require too much disk space (I have been in situations where the chi2 grid would be as large as several TB!). Usually, one is not interested in the chi2 of *all* models, but only those that match the data within some tolerance threshold. This option allows you to only save on the disk the models that are worst than the best chi2 by some amount ```chi2 - best_chi2 < SAVE_BESTCHI``` (where ```SAVE_BESTCHI=1``` if you are interested in standard 68% confidence intervals, or ```2.71``` for 90% confidence, etc., see Avni 1976). If you set the ```INTERVAL_FROM_CHI2``` option, the program will use these saved grids to automatically compute parameter confidence intervals. The parameters of all these "good" models are saved in a separate ".grid" file for each galaxy of the input catalog, inside the ```best_chi2``` folder. The format is similar to the ".grid" file for the full chi2 grid (which is described above), but not identical. The ``fast++-grid2fits`` tool can also convert these files into FITS tables. The binary format is the following:
0 commit comments