Skip to content

Commit

Permalink
Merge branch 'master' into fix-project-switcher-path
Browse files Browse the repository at this point in the history
  • Loading branch information
@patricio-posthog
patricio-posthog authored Oct 17, 2024
2 parents 3e95539 + 1c4f46a commit 0d95282
Show file tree
Hide file tree
Showing 25 changed files with 1,070 additions and 516 deletions.
2 changes: 1 addition & 1 deletion ee/clickhouse/queries/experiments/funnel_experiment_result.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@

from posthog.constants import ExperimentNoResultsErrorKeys
from posthog.hogql_queries.experiments import CONTROL_VARIANT_KEY
from posthog.hogql_queries.experiments.funnel_statistics import (
from posthog.hogql_queries.experiments.funnels_statistics import (
are_results_significant,
calculate_credible_intervals,
calculate_probabilities,
Expand Down
88 changes: 44 additions & 44 deletions ee/clickhouse/queries/experiments/test_funnel_experiment_result.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4,13 +4,13 @@

from flaky import flaky

from posthog.hogql_queries.experiments.funnel_statistics import (
from posthog.hogql_queries.experiments.funnels_statistics import (
are_results_significant,
calculate_expected_loss,
calculate_probabilities,
calculate_credible_intervals as calculate_funnel_credible_intervals,
)
from posthog.schema import ExperimentSignificanceCode, ExperimentVariantFunnelResult
from posthog.schema import ExperimentSignificanceCode, ExperimentVariantFunnelsBaseStats

Probability = float

Expand All @@ -25,7 +25,7 @@ def logbeta(x: int, y: int) -> float:


def calculate_probability_of_winning_for_target(
target_variant: ExperimentVariantFunnelResult, other_variants: list[ExperimentVariantFunnelResult]
target_variant: ExperimentVariantFunnelsBaseStats, other_variants: list[ExperimentVariantFunnelsBaseStats]
) -> Probability:
"""
Calculates the probability of winning for target variant.
Expand Down Expand Up @@ -146,8 +146,8 @@ def probability_D_beats_A_B_and_C(
@flaky(max_runs=10, min_passes=1)
class TestFunnelExperimentCalculator(unittest.TestCase):
def test_calculate_results(self):
variant_test = ExperimentVariantFunnelResult(key="A", success_count=100, failure_count=10)
variant_control = ExperimentVariantFunnelResult(key="B", success_count=100, failure_count=18)
variant_test = ExperimentVariantFunnelsBaseStats(key="A", success_count=100, failure_count=10)
variant_control = ExperimentVariantFunnelsBaseStats(key="B", success_count=100, failure_count=18)

_, probability = calculate_probabilities(variant_control, [variant_test])
self.assertAlmostEqual(probability, 0.918, places=2)
Expand All @@ -164,8 +164,8 @@ def test_calculate_results(self):
self.assertAlmostEqual(credible_intervals[variant_test.key][1], 0.9494, places=3)

def test_simulation_result_is_close_to_closed_form_solution(self):
variant_test = ExperimentVariantFunnelResult(key="A", success_count=100, failure_count=10)
variant_control = ExperimentVariantFunnelResult(key="B", success_count=100, failure_count=18)
variant_test = ExperimentVariantFunnelsBaseStats(key="A", success_count=100, failure_count=10)
variant_control = ExperimentVariantFunnelsBaseStats(key="B", success_count=100, failure_count=18)

_, probability = calculate_probabilities(variant_control, [variant_test])
self.assertAlmostEqual(probability, 0.918, places=1)
Expand All @@ -174,9 +174,9 @@ def test_simulation_result_is_close_to_closed_form_solution(self):
self.assertAlmostEqual(probability, alternative_probability, places=1)

def test_calculate_results_for_two_test_variants(self):
variant_test_1 = ExperimentVariantFunnelResult(key="A", success_count=100, failure_count=10)
variant_test_2 = ExperimentVariantFunnelResult(key="B", success_count=100, failure_count=3)
variant_control = ExperimentVariantFunnelResult(key="C", success_count=100, failure_count=18)
variant_test_1 = ExperimentVariantFunnelsBaseStats(key="A", success_count=100, failure_count=10)
variant_test_2 = ExperimentVariantFunnelsBaseStats(key="B", success_count=100, failure_count=3)
variant_control = ExperimentVariantFunnelsBaseStats(key="C", success_count=100, failure_count=18)

probabilities = calculate_probabilities(variant_control, [variant_test_1, variant_test_2])
self.assertAlmostEqual(sum(probabilities), 1)
Expand Down Expand Up @@ -210,9 +210,9 @@ def test_calculate_results_for_two_test_variants(self):
self.assertAlmostEqual(credible_intervals[variant_test_2.key][1], 0.9894, places=3)

def test_calculate_results_for_two_test_variants_almost_equal(self):
variant_test_1 = ExperimentVariantFunnelResult(key="A", success_count=120, failure_count=60)
variant_test_2 = ExperimentVariantFunnelResult(key="B", success_count=110, failure_count=52)
variant_control = ExperimentVariantFunnelResult(key="C", success_count=130, failure_count=65)
variant_test_1 = ExperimentVariantFunnelsBaseStats(key="A", success_count=120, failure_count=60)
variant_test_2 = ExperimentVariantFunnelsBaseStats(key="B", success_count=110, failure_count=52)
variant_control = ExperimentVariantFunnelsBaseStats(key="C", success_count=130, failure_count=65)

probabilities = calculate_probabilities(variant_control, [variant_test_1, variant_test_2])
self.assertAlmostEqual(sum(probabilities), 1)
Expand Down Expand Up @@ -245,8 +245,8 @@ def test_calculate_results_for_two_test_variants_almost_equal(self):
self.assertAlmostEqual(credible_intervals[variant_test_2.key][1], 0.7460, places=3)

def test_absolute_loss_less_than_one_percent_but_not_significant(self):
variant_test_1 = ExperimentVariantFunnelResult(key="A", success_count=286, failure_count=2014)
variant_control = ExperimentVariantFunnelResult(key="B", success_count=267, failure_count=2031)
variant_test_1 = ExperimentVariantFunnelsBaseStats(key="A", success_count=286, failure_count=2014)
variant_control = ExperimentVariantFunnelsBaseStats(key="B", success_count=267, failure_count=2031)

probabilities = calculate_probabilities(variant_control, [variant_test_1])
self.assertAlmostEqual(sum(probabilities), 1)
Expand All @@ -267,10 +267,10 @@ def test_absolute_loss_less_than_one_percent_but_not_significant(self):
self.assertAlmostEqual(credible_intervals[variant_test_1.key][1], 0.1384, places=3)

def test_calculate_results_for_three_test_variants(self):
variant_test_1 = ExperimentVariantFunnelResult(key="A", success_count=100, failure_count=10)
variant_test_2 = ExperimentVariantFunnelResult(key="B", success_count=100, failure_count=3)
variant_test_3 = ExperimentVariantFunnelResult(key="C", success_count=100, failure_count=30)
variant_control = ExperimentVariantFunnelResult(key="D", success_count=100, failure_count=18)
variant_test_1 = ExperimentVariantFunnelsBaseStats(key="A", success_count=100, failure_count=10)
variant_test_2 = ExperimentVariantFunnelsBaseStats(key="B", success_count=100, failure_count=3)
variant_test_3 = ExperimentVariantFunnelsBaseStats(key="C", success_count=100, failure_count=30)
variant_control = ExperimentVariantFunnelsBaseStats(key="D", success_count=100, failure_count=18)

probabilities = calculate_probabilities(variant_control, [variant_test_1, variant_test_2, variant_test_3])
self.assertAlmostEqual(sum(probabilities), 1)
Expand Down Expand Up @@ -313,10 +313,10 @@ def test_calculate_results_for_three_test_variants(self):
self.assertAlmostEqual(credible_intervals[variant_test_3.key][1], 0.8332, places=3)

def test_calculate_results_for_three_test_variants_almost_equal(self):
variant_test_1 = ExperimentVariantFunnelResult(key="A", success_count=120, failure_count=60)
variant_test_2 = ExperimentVariantFunnelResult(key="B", success_count=110, failure_count=52)
variant_test_3 = ExperimentVariantFunnelResult(key="C", success_count=100, failure_count=46)
variant_control = ExperimentVariantFunnelResult(key="D", success_count=130, failure_count=65)
variant_test_1 = ExperimentVariantFunnelsBaseStats(key="A", success_count=120, failure_count=60)
variant_test_2 = ExperimentVariantFunnelsBaseStats(key="B", success_count=110, failure_count=52)
variant_test_3 = ExperimentVariantFunnelsBaseStats(key="C", success_count=100, failure_count=46)
variant_control = ExperimentVariantFunnelsBaseStats(key="D", success_count=130, failure_count=65)

probabilities = calculate_probabilities(variant_control, [variant_test_1, variant_test_2, variant_test_3])
self.assertAlmostEqual(sum(probabilities), 1)
Expand Down Expand Up @@ -357,10 +357,10 @@ def test_calculate_results_for_three_test_variants_almost_equal(self):
self.assertAlmostEqual(credible_intervals[variant_test_3.key][1], 0.7547, places=3)

def test_calculate_results_for_three_test_variants_much_better_than_control(self):
variant_test_1 = ExperimentVariantFunnelResult(key="A", success_count=130, failure_count=60)
variant_test_2 = ExperimentVariantFunnelResult(key="B", success_count=135, failure_count=62)
variant_test_3 = ExperimentVariantFunnelResult(key="C", success_count=132, failure_count=60)
variant_control = ExperimentVariantFunnelResult(key="D", success_count=80, failure_count=65)
variant_test_1 = ExperimentVariantFunnelsBaseStats(key="A", success_count=130, failure_count=60)
variant_test_2 = ExperimentVariantFunnelsBaseStats(key="B", success_count=135, failure_count=62)
variant_test_3 = ExperimentVariantFunnelsBaseStats(key="C", success_count=132, failure_count=60)
variant_control = ExperimentVariantFunnelsBaseStats(key="D", success_count=80, failure_count=65)

probabilities = calculate_probabilities(variant_control, [variant_test_1, variant_test_2, variant_test_3])
self.assertAlmostEqual(sum(probabilities), 1)
Expand Down Expand Up @@ -392,14 +392,14 @@ def test_calculate_results_for_three_test_variants_much_better_than_control(self
self.assertAlmostEqual(credible_intervals[variant_test_3.key][1], 0.7488, places=3)

def test_calculate_results_for_seven_test_variants(self):
variant_test_1 = ExperimentVariantFunnelResult(key="A", success_count=100, failure_count=17)
variant_test_2 = ExperimentVariantFunnelResult(key="B", success_count=100, failure_count=16)
variant_test_3 = ExperimentVariantFunnelResult(key="C", success_count=100, failure_count=30)
variant_test_4 = ExperimentVariantFunnelResult(key="D", success_count=100, failure_count=31)
variant_test_5 = ExperimentVariantFunnelResult(key="E", success_count=100, failure_count=29)
variant_test_6 = ExperimentVariantFunnelResult(key="F", success_count=100, failure_count=32)
variant_test_7 = ExperimentVariantFunnelResult(key="G", success_count=100, failure_count=33)
variant_control = ExperimentVariantFunnelResult(key="H", success_count=100, failure_count=18)
variant_test_1 = ExperimentVariantFunnelsBaseStats(key="A", success_count=100, failure_count=17)
variant_test_2 = ExperimentVariantFunnelsBaseStats(key="B", success_count=100, failure_count=16)
variant_test_3 = ExperimentVariantFunnelsBaseStats(key="C", success_count=100, failure_count=30)
variant_test_4 = ExperimentVariantFunnelsBaseStats(key="D", success_count=100, failure_count=31)
variant_test_5 = ExperimentVariantFunnelsBaseStats(key="E", success_count=100, failure_count=29)
variant_test_6 = ExperimentVariantFunnelsBaseStats(key="F", success_count=100, failure_count=32)
variant_test_7 = ExperimentVariantFunnelsBaseStats(key="G", success_count=100, failure_count=33)
variant_control = ExperimentVariantFunnelsBaseStats(key="H", success_count=100, failure_count=18)

probabilities = calculate_probabilities(
variant_control,
Expand Down Expand Up @@ -487,8 +487,8 @@ def test_calculate_results_for_seven_test_variants(self):
self.assertAlmostEqual(credible_intervals[variant_test_7.key][1], 0.8174, places=3)

def test_calculate_results_control_is_significant(self):
variant_test = ExperimentVariantFunnelResult(key="test", success_count=100, failure_count=18)
variant_control = ExperimentVariantFunnelResult(key="control", success_count=100, failure_count=10)
variant_test = ExperimentVariantFunnelsBaseStats(key="test", success_count=100, failure_count=18)
variant_control = ExperimentVariantFunnelsBaseStats(key="control", success_count=100, failure_count=10)

probabilities = calculate_probabilities(variant_control, [variant_test])

Expand All @@ -507,13 +507,13 @@ def test_calculate_results_control_is_significant(self):
self.assertAlmostEqual(credible_intervals[variant_test.key][1], 0.9010, places=3)

def test_calculate_results_many_variants_control_is_significant(self):
variant_test_1 = ExperimentVariantFunnelResult(key="test_1", success_count=100, failure_count=20)
variant_test_2 = ExperimentVariantFunnelResult(key="test_2", success_count=100, failure_count=21)
variant_test_3 = ExperimentVariantFunnelResult(key="test_3", success_count=100, failure_count=22)
variant_test_4 = ExperimentVariantFunnelResult(key="test_4", success_count=100, failure_count=23)
variant_test_5 = ExperimentVariantFunnelResult(key="test_5", success_count=100, failure_count=24)
variant_test_6 = ExperimentVariantFunnelResult(key="test_6", success_count=100, failure_count=25)
variant_control = ExperimentVariantFunnelResult(key="control", success_count=100, failure_count=10)
variant_test_1 = ExperimentVariantFunnelsBaseStats(key="test_1", success_count=100, failure_count=20)
variant_test_2 = ExperimentVariantFunnelsBaseStats(key="test_2", success_count=100, failure_count=21)
variant_test_3 = ExperimentVariantFunnelsBaseStats(key="test_3", success_count=100, failure_count=22)
variant_test_4 = ExperimentVariantFunnelsBaseStats(key="test_4", success_count=100, failure_count=23)
variant_test_5 = ExperimentVariantFunnelsBaseStats(key="test_5", success_count=100, failure_count=24)
variant_test_6 = ExperimentVariantFunnelsBaseStats(key="test_6", success_count=100, failure_count=25)
variant_control = ExperimentVariantFunnelsBaseStats(key="control", success_count=100, failure_count=10)

variants_test = [
variant_test_1,
Expand Down
Loading

0 comments on commit 0d95282

Please sign in to comment.