JP-3658: JWST Accommodation of the Ramp Fitting Likelihood Algorithm

CHANGES.rst

@@ -187,14 +187,17 @@ pipeline
 
 - Updated ``calwebb_spec2`` to run ``nsclean`` on NIRSpec imprint and background 
   association members. [#8786]
-
+
+- Updated `calwebb_spec3` to not save the `pixel_replacement` output by default.[#8765]
+
 ramp_fitting
 ------------
 
 - Moved the read noise variance recalculation for CHARGELOSS flagging to the C
   implementation, when the algorithm is ``OLS_C``. [#8697, spacetelescope/stcal#278]
 
-- Updated `calwebb_spec3` to not save the `pixel_replacement` output by default.[#8765] 
+- Updated the flow of the detector 1 pipeline when selecting the ``LIKELY`` algorithm
+  for ramp fitting.  The ramps must contain a minimum number of groups (4).[#8631]
 
 resample
 --------

docs/jwst/ramp_fitting/description.rst

@@ -54,6 +54,11 @@ JWST-specific features are described later within this document.
 Upon successful completion of this step, the status keyword S_RAMP will be set
 to "COMPLETE".
 
+There is a new algorithm available for testing in ramp fitting, which is the
+likelihood algorithm.  It is selected by setting ``--ramp_fitting.algorithm=LIKELY``.
+The details are in the ``stcal`` documentation at
+:ref:`Likelihood Algorithm Details <stcal:likelihood_algo>`.
+
 Multiprocessing
 ---------------
 This step has the option of running in multiprocessing mode. In that mode it will

jwst/ramp_fitting/ramp_fit_step.py

@@ -5,6 +5,8 @@
 from stcal.ramp_fitting import ramp_fit
 from stcal.ramp_fitting import utils
 
+from stcal.ramp_fitting.likely_fit import LIKELY_MIN_NGROUPS
+
 from stcal.ramp_fitting.utils import LARGE_VARIANCE
 from stcal.ramp_fitting.utils import LARGE_VARIANCE_THRESHOLD
 
@@ -385,7 +387,7 @@ class RampFitStep(Step):
     class_alias = "ramp_fit"
 
     spec = """
-        algorithm = option('OLS', 'OLS_C', default='OLS_C') # 'OLS' and 'OLS_C' use the same underlying algorithm, but OLS_C is implemented in C
+        algorithm = option('OLS', 'OLS_C', 'LIKELY', default='OLS_C') # 'OLS' and 'OLS_C' use the same underlying algorithm, but OLS_C is implemented in C
         int_name = string(default='')
         save_opt = boolean(default=False) # Save optional output
         opt_name = string(default='')
@@ -414,8 +416,16 @@ def process(self, input):
             readnoise_filename = self.get_reference_file(input_model, 'readnoise')
             gain_filename = self.get_reference_file(input_model, 'gain')
 
-            log.info('Using READNOISE reference file: %s', readnoise_filename)
-            log.info('Using GAIN reference file: %s', gain_filename)
+            ngroups = input_model.data.shape[1]
+            if self.algorithm.upper() == "LIKELY" and ngroups < LIKELY_MIN_NGROUPS:
+                log.info(f"When selecting the LIKELY ramp fitting algorithm the"
+                         f" ngroups needs to be a minimum of {LIKELY_MIN_NGROUPS},"
+                         f" but ngroups = {ngroups}.  Due to this, the ramp fitting algorithm"
+                         f" is being changed to OLS_C")
+                self.algorithm = "OLS_C"
+
+            log.info(f"Using READNOISE reference file: {readnoise_filename}")
+            log.info(f"Using GAIN reference file: {gain_filename}")
 
             with datamodels.ReadnoiseModel(readnoise_filename) as readnoise_model, \
                     datamodels.GainModel(gain_filename) as gain_model:
@@ -432,8 +442,8 @@ def process(self, input):
                 readnoise_2d, gain_2d = get_reference_file_subarrays(
                     input_model, readnoise_model, gain_model, frames_per_group)
 
-            log.info('Using algorithm = %s' % self.algorithm)
-            log.info('Using weighting = %s' % self.weighting)
+            log.info(f"Using algorithm = {self.algorithm}")
+            log.info(f"Using weighting = {self.weighting}")
 
             buffsize = ramp_fit.BUFSIZE
             if self.algorithm == "GLS":