Merge pull request #12 from transferwise/linting

linting round 3

.flake8

@@ -6,4 +6,12 @@ exclude =
   docs
   build
   dist
-  venv
+  venv
+
+per-file-ignores =
+  tw_experimentation/streamlit/pages_wrap/page4_Experiment_Evaluation.py:E501
+  tw_experimentation/streamlit/pages_wrap/page3_Monitoring.py:E501
+  tw_experimentation/streamlit/pages_wrap/page2_Experiment_Design.py:E501
+  tw_experimentation/streamlit/pages_wrap/page1_Data_Loading.py:E501
+  tw_experimentation/streamlit/pages_wrap/page5_Experiment_Evaluation_Bayesian.py:E501
+  tw_experimentation/streamlit/Main.py:E501

notebooks/2_integrity_checks + evaluation.ipynb

@@ -29,21 +29,18 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "import sys  \n",
+    "import sys\n",
     "import os\n",
     "root_path = os.path.realpath('..')\n",
     "sys.path.insert(0, root_path)\n",
     "\n",
-    "from tw_experimentation.utils import ExperimentDataset, variantname_color_map, variant_name_map\n",
+    "from tw_experimentation.utils import ExperimentDataset\n",
     "from tw_experimentation.data_generation import RevenueConversion\n",
     "from tw_experimentation.widgetizer import FrequentistEvaluation, SegmentationInterface, Segmentation\n",
     "from tw_experimentation.statistical_tests import cuped\n",
     "\n",
-    "import pandas as pd\n",
     "import numpy as np\n",
     "import pickle\n",
-    "%matplotlib inline\n",
-    "import matplotlib.pyplot as plt\n",
     "\n",
     "import plotly.io as pio\n",
     "pio.renderers.default = \"png\"\n",
@@ -93,7 +90,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -104,14 +101,14 @@
     "\n",
     "rc = RevenueConversion()\n",
     "\n",
-    "df_abn = rc.generate_data_abn_test(  \n",
+    "df_abn = rc.generate_data_abn_test(\n",
     "    n=10000,\n",
     "    n_treatments = 2,\n",
     "    baseline_conversion = .4,\n",
-    "    treatment_effect_conversion=.04,\n",
+    "    treatment_effect_conversion=.02,\n",
     "    baseline_mean_revenue = 6,\n",
     "    sigma_revenue = 1,\n",
-    "    treatment_effect_revenue = .1 \n",
+    "    treatment_effect_revenue = .05\n",
     ")\n",
     "\n",
     "df_abn['currency'] = np.random.choice(['GBP', 'EUR', 'USD'], size=len(df_abn))\n",
@@ -235,7 +232,7 @@
     "# ]\n",
     "\n",
     "# metrics = [ # Specify metrics types, or leave empty to infer automatically\n",
-    "#     'binary', \n",
+    "#     'binary',\n",
     "#     'continuous'\n",
     "#     ]\n",
     "\n",
@@ -1167,7 +1164,7 @@
     "segmentation = Segmentation(ed=experiment_dataset)\n",
     "wise_pizza_segmentation = segmentation.wise_pizza_frequentist(\n",
     "    target='revenue',\n",
-    "    treatment=treatment, \n",
+    "    treatment=treatment,\n",
     "    segments=segments,\n",
     "    max_depth=2, # customize to own preference\n",
     "    min_segments=3 # customize to own preference\n",

tests/test_bayes_in_notebook.py

@@ -44,24 +44,27 @@ def test_bayestest_end(experiment_data_two_treatment):
     br.fig_posterior_difference_by_target(outcome_metric)
 
     # greater than zero probability
-    description = f"The probability that the average treatment effect is greater than 0 for {outcome_metric} is\n"
+    description = f"The probability that the average treatment effect is greater than 0 for {outcome_metric} is\n"  # noqa: E501
     for variant in range(1, ed.n_variants):
-        description += f" {br.prob_greater_than_zero(outcome_metric)[variant]*100}% for variant {variant}"
+        description += f" {br.prob_greater_than_zero(outcome_metric)[variant]*100}% for variant {variant}"  # noqa: E501
         description += "." if variant == ed.n_variants - 1 else ",\n"
     print(description)
 
     # greater than threshold probability
     threshold = 100
-    description = f"The probability that the average treatment effect is greater than {threshold} for {outcome_metric} is\n"
+    description = "The probability that the average treatment effect is greater than "
+    f"{threshold} for {outcome_metric} is\n"
     for variant in range(1, ed.n_variants):
-        description += f" {br.prob_greater_than_z(threshold, outcome_metric)[variant]*100}% for variant {variant}"
+        description += f" {br.prob_greater_than_z(threshold, outcome_metric)[variant]*100}% for variant {variant}"  # noqa: E501
         description += "." if variant == ed.n_variants - 1 else ",\n"
     print(description)
 
     # rope probability
     probs, rope_lower, rope_upper = br.rope(outcome_metric)
 
-    description = f"The probability that the average treatment effect is outside the region of practical equivalence ({rope_lower},{rope_upper}) for {outcome_metric} is\n"
+    description = "The probability that the average treatment effect is outside"
+    " the region of practical equivalence"
+    f" ({rope_lower},{rope_upper}) for {outcome_metric} is\n"
     for variant in range(1, ed.n_variants):
         description += f" {probs[variant]*100:.2f}% for variant {variant}"
         description += "." if variant == ed.n_variants - 1 else ",\n"

tests/test_bayestest.py

@@ -1,13 +1,10 @@
 import pytest
-import warnings
 
 from tw_experimentation.utils import ExperimentDataset
 
 from tw_experimentation.bayes.bayes_test import BayesTest
 from tw_experimentation.data_generation import RevenueConversion
 
-from numpyro.distributions import LogNormal, Gamma, Uniform
-
 
 class TestBayesTesterEndToEnd(object):
     def test_with_new_model(self):

tests/test_datasets.py

@@ -43,7 +43,8 @@ def test_check_preprocessor(experiment_data):
     """checks if dataset can be preprocessed (dummy encoding of vars etc...)
 
     Args:
-        df (pd.DataFrame): dataset for experiments, with cols for treatment, targets and covariates
+        df (pd.DataFrame): dataset for experiments, with cols for treatment,
+            targets and covariates
     """
     ed = experiment_data
     ed.preprocess_dataset()

tests/test_experiment_design.py

@@ -23,10 +23,12 @@ def experiment_data():
 
 
 def test_check_preprocessor(experiment_data):
-    """checks if dataset can be preprocessed (dummy encoding of vars etc...)
+    """checks if dataset can be preprocessed
+        (dummy encoding of vars etc...)
 
     Args:
-        df (pd.DataFrame): dataset for experiments, with cols for treatment, targets and covariates
+        df (pd.DataFrame): dataset for experiments, with cols
+            for treatment, targets and covariates
     """
     ed = experiment_data
     ed.preprocess_pre_experiment_dataset()
@@ -58,7 +60,7 @@ def test_auto_sample_size_filler(experiment_data):
                 relation="relative",
                 treatment_share=0.5,
             )
-            sample_size = setup.sample_size_two_sample_proportion_z_test()
+            _ = setup.sample_size_two_sample_proportion_z_test()
         elif exp_design_metric_type == "continuous":
             setup = Setuper.from_uplift(
                 alpha=0.1,
@@ -69,7 +71,7 @@ def test_auto_sample_size_filler(experiment_data):
                 relation="relative",
                 treatment_share=0.4,
             )
-            sample_size = setup.sample_size_t_test()
+            setup.sample_size_t_test()
         else:
             raise ValueError("metric type not supported")
 

tests/test_frequentist_hypothesis_tests.py

@@ -1,7 +1,5 @@
 import pytest
 
-from pandas import testing as tm
-
 from tw_experimentation.utils import ExperimentDataset
 from tw_experimentation.data_generation import RevenueConversion
 
@@ -10,6 +8,8 @@
     compute_frequentist_results,
 )
 
+from pandas import testing as tm
+
 
 @pytest.fixture
 def experiment_data_continuous(request):

tests/test_frequentist_static.py

@@ -6,10 +6,9 @@
 from tw_experimentation.data_generation import RevenueConversion
 
 from tw_experimentation.statistical_tests import FrequentistTest, BaseTest
+
 from tw_experimentation.widgetizer import (
     FrequentistEvaluation,
-    Segmentation,
-    SegmentationInterface,
 )
 
 from tw_experimentation.statistical_tests import cuped
@@ -142,30 +141,5 @@ def test_frequentist_in_notebook(experiment_data_two_treatment):
     cuped(experiment_data_two_treatment, has_correction="Yes", alpha=0.05)
 
 
-def test_frequentist_wise_pizza_segmentation_in_notebook(experiment_data_with_segments):
-    segments = ["segment_1", "segment_2", "segment_3", "currency"]
-
-    # change to treatment number of interest
-    treatment = 1
-
-    segmentation = Segmentation(ed=experiment_data_with_segments)
-    wise_pizza_segmentation = segmentation.wise_pizza_frequentist(
-        target="conversion",
-        treatment=treatment,
-        segments=segments,
-        max_depth=2,  # customize to own preference
-        min_segments=10,  # customize to own preference
-    )
-
-    slice_finder = segmentation.get_sf()
-    slice_finder.plot(width=1000, plot_is_static=True)
-
-
-def test_frequentist_custom_segmentation_in_notebook(experiment_data_with_segments):
-    segments = ["segment_1", "segment_2", "segment_3", "currency"]
-
-    _ = SegmentationInterface(experiment_data_with_segments)
-
-
 if __name__ == "__main__":
     pytest.main([__file__])

tests/test_integrity_checks_in_notebook.py

@@ -1,6 +1,5 @@
 import pytest
 
-import pandas as pd
 import numpy as np
 from tw_experimentation.utils import ExperimentDataset
 from tw_experimentation.data_generation import RevenueConversion
@@ -61,13 +60,13 @@ def test_monitoring_segments_in_notebook(experiment_data_with_segments):
     # choose segments to monitor
     segments = ["currency", "segment_1", "segment_2"]
     segment_monitor = SegmentMonitoring(ed, segments)
-    segment_monitor_results = segment_monitor.create_tables_and_plots()
+    _ = segment_monitor.create_tables_and_plots()
 
 
 def test_monitoring_in_notebook(experiment_data_with_segments):
     ed = experiment_data_with_segments
     monitor = Monitoring(ed)
-    monitor_results = monitor.create_tables_and_plots()
+    _ = monitor.create_tables_and_plots()
 
 
 def test_normality_checks(experiment_data_with_segments):
@@ -84,8 +83,8 @@ def test_sequential_test_in_notebook(experiment_data_with_segments):
     ed = experiment_data_with_segments
     if ed.is_dynamic_observation:
         sequential_test = SequentialTest(ed)
-        sequential_test_results = sequential_test.sequential_test_results()
-        fig = sequential_test.fig_sequential_test()
+        sequential_test.sequential_test_results()
+        _ = sequential_test.fig_sequential_test()
 
 
 if __name__ == "__main__":

tests/test_numpyro_sampler_model.py

@@ -11,9 +11,6 @@
 from numpyro.infer import MCMC, NUTS
 
 
-from tw_experimentation.statistical_tests import *
-
-
 class TestNumpyroSampler(object):
     @pytest.fixture
     def data_input(self):