add comments from #109

mesmer/calibrate_mesmer/train_lt.py

@@ -96,6 +96,7 @@ def train_lt(preds, targs, esm, cfg, save_params=True):
 
     # specify necessary variables from config file
     wgt_scen_tr_eq = cfg.wgt_scen_tr_eq
+    # This code will ever only work with a single target variable
     method_lt = cfg.methods[targ_name]["lt"]
     method_lv = cfg.methods[targ_name]["lv"]
     method_lt_each_gp_sep = cfg.method_lt_each_gp_sep
@@ -182,6 +183,8 @@ def train_lt(preds, targs, esm, cfg, save_params=True):
 
             for pred in params_lv["preds"]:
                 params_lv[f"coef_{pred}"][targ] = reg_xr[pred].values
+    else:
+        raise NotImplementedError()
 
     # save the local trend paramters if requested
     if save_params:

mesmer/calibrate_mesmer/train_lv.py

@@ -249,6 +249,12 @@ def train_lv_AR1_sci(params_lv, targs, y, wgt_scen_eq, aux, cfg):
         params_lv["AR1_std_innovs"][targ_name] = params_scen.standard_deviation.values
 
         # determine localization radius, empirical cov matrix, and localized ecov matrix
+
+        # y.dims = (sample, gridpoint)
+        # wgt_scen_eq.dims = (sample,)
+        # aux["phi_gc"].dims = (gridpoint, gripoint)
+        # where sample = is a stacked "time, scenario, ensmember"
+
         res = train_lv_find_localized_ecov(y[targ_name], wgt_scen_eq, aux, cfg)
         params_lv["L"][targ_name] = res.localization_radius.values
         params_lv["ecov"][targ_name] = res.covariance.values

mesmer/create_emulations/create_emus_gv.py

@@ -151,6 +151,7 @@ def create_emus_gv_AR(params_gv, nr_emus_v, nr_ts_emus_v, seed):
     np.random.seed(seed)
 
     # buffer so that initial start at 0 does not influence overall result
+    # Should this buffer be based on the length of ar_lags instead of hard-coded?
     buffer = 50
 
     # re-name params for easier reading of code below