Denominator Forecasting

notebooks/hosp_only_ww_model.qmd

@@ -35,7 +35,7 @@ We begin by loading the Stan data, converting it the correct inputs for our mode
 
 ```{python}
 # | label: create model
-my_hosp_only_ww_model, data_observed_hospital_admissions = (
+my_hosp_only_ww_model, data_observed_disease_hospital_admissions = (
     create_hosp_only_ww_model_from_stan_data(
         "data/fit_hosp_only/stan_data.json"
     )
@@ -50,7 +50,7 @@ We check that we can simulate from the prior predictive
 n_forecast_days = 35
 
 prior_predictive = my_hosp_only_ww_model.prior_predictive(
-    n_datapoints=len(data_observed_hospital_admissions) + n_forecast_days,
+    n_datapoints=len(data_observed_disease_hospital_admissions) + n_forecast_days,
     numpyro_predictive_args={"num_samples": 200},
 )
 ```
@@ -64,7 +64,7 @@ my_hosp_only_ww_model.run(
     num_warmup=500,
     num_samples=500,
     rng_key=jax.random.key(200),
-    data_observed_hospital_admissions=data_observed_hospital_admissions,
+    data_observed_disease_hospital_admissions=data_observed_disease_hospital_admissions,
     mcmc_args=dict(num_chains=4, progress_bar=False),
     nuts_args=dict(find_heuristic_step_size=True),
 )
@@ -75,7 +75,7 @@ Create the posterior predictive and forecast:
 ```{python}
 # | label: posterior predictive
 posterior_predictive = my_hosp_only_ww_model.posterior_predictive(
-    n_datapoints=len(data_observed_hospital_admissions) + n_forecast_days
+    n_datapoints=len(data_observed_disease_hospital_admissions) + n_forecast_days
 )
 ```
 

nssp_demo/build_model.py

@@ -45,8 +45,8 @@ def build_model_from_dir(model_dir):
         jnp.array(model_data["generation_interval_pmf"]),
     )  # check if off by 1 or reversed
 
-    data_observed_hospital_admissions = jnp.array(
-        model_data["data_observed_hospital_admissions"]
+    data_observed_disease_hospital_admissions = jnp.array(
+        model_data["data_observed_disease_hospital_admissions"]
     )
     state_pop = jnp.array(model_data["state_pop"])
 
@@ -84,4 +84,8 @@ def build_model_from_dir(model_dir):
         n_initialization_points=uot,
     )
 
-    return my_model, data_observed_hospital_admissions, right_truncation_offset
+    return (
+        my_model,
+        data_observed_disease_hospital_admissions,
+        right_truncation_offset,
+    )

nssp_demo/fit_model.py

@@ -21,14 +21,16 @@
 args = parser.parse_args()
 model_dir = args.model_dir
 
-my_model, data_observed_hospital_admissions, right_truncation_offset = (
-    build_model_from_dir(model_dir)
-)
+(
+    my_model,
+    data_observed_disease_hospital_admissions,
+    right_truncation_offset,
+) = build_model_from_dir(model_dir)
 my_model.run(
     num_warmup=500,
     num_samples=500,
     rng_key=jax.random.key(200),
-    data_observed_hospital_admissions=data_observed_hospital_admissions,
+    data_observed_disease_hospital_admissions=data_observed_disease_hospital_admissions,
     right_truncation_offset=right_truncation_offset,
     mcmc_args=dict(num_chains=n_chains, progress_bar=True),
     nuts_args=dict(find_heuristic_step_size=True),

nssp_demo/generate_predictive.py

@@ -27,9 +27,11 @@
 args = parser.parse_args()
 model_dir = args.model_dir
 n_forecast_points = args.n_forecast_points
-my_model, data_observed_hospital_admissions, right_truncation_offset = (
-    build_model_from_dir(model_dir)
-)
+(
+    my_model,
+    data_observed_disease_hospital_admissions,
+    right_truncation_offset,
+) = build_model_from_dir(model_dir)
 
 my_model._init_model(1, 1)
 fresh_sampler = my_model.mcmc.sampler
@@ -47,12 +49,13 @@
 #         "num_samples": my_model.mcmc.num_samples * my_model.mcmc.num_chains,
 #         "batch_ndims":1
 #     },
-#     n_datapoints=len(data_observed_hospital_admissions) + n_forecast_points,
+#     n_datapoints=len(data_observed_disease_hospital_admissions) + n_forecast_points,
 # )
 # need to figure out a way to generate these as distinct chains, so that the result of the to_datarame method is more compact
 
 posterior_predictive = my_model.posterior_predictive(
-    n_datapoints=len(data_observed_hospital_admissions) + n_forecast_points
+    n_datapoints=len(data_observed_disease_hospital_admissions)
+    + n_forecast_points
 )
 
 idata = az.from_numpyro(