algorithms/get_results_mi.py

"""
Analyze Results from MI PRR
Copyright (c) 2023 Cannlytics

Authors: Keegan Skeate <https://github.com/keeganskeate>
Created: 10/23/2023
Updated: 7/11/2024
License: MIT License <https://github.com/cannlytics/cannabis-data-science/blob/main/LICENSE>

Data Sources:
    
    - Public records request

"""
# External imports:
from datetime import datetime
import os
import matplotlib.pyplot as plt
from matplotlib.ticker import StrMethodFormatter
from matplotlib import cm
import numpy as np
import pandas as pd
import re
import seaborn as sns
from scipy import stats


# Setup plotting style.
plt.style.use('fivethirtyeight')
plt.rcParams.update({
    'figure.figsize': (12, 8),
    'font.family': 'Times New Roman',
    'font.size': 24,
})


def save_figure(filename, dpi=300, bbox_inches='tight'):
    """Save a figure to the figures directory."""
    plt.savefig(f'figures/{filename}', bbox_inches=bbox_inches, dpi=dpi)


# === Get the data ===

# Read the results.
data_dir = r'D:\data\public-records\Michigan'
datafile = os.path.join(data_dir, 'Michigan_Metrc_Flower_Potency_Final_2.17.23.xlsx')
mi_results = pd.read_excel(datafile)


# === Clean the data ===

# Rename certain columns.
mi_results = mi_results.rename(columns={
    'ProductName': 'product_name',
    'ProductCategory': 'product_type',
    'TestType': 'test_type',
    'Quantity': 'total_thc',
    'Licensee': 'lab',
    'TestPerformedDate': 'date_tested',
    'Comment': 'notes',
    'Med AU': 'medical',
})

# Standardize state.
state = 'MI'
mi_results['lab_state'] = state
mi_results['producer_state'] = state

# Add a date column.
mi_results['date'] = pd.to_datetime(mi_results['date_tested'], format='mixed')
mi_results['week'] = mi_results['date'].dt.to_period('W').astype(str)
mi_results['month'] = mi_results['date'].dt.to_period('M').astype(str)
mi_results = mi_results.sort_values('date')

# Save the results.
outfile = 'D://data/michigan/mi-results-latest.xlsx'
outfile_csv = 'D://data/michigan/mi-results-latest.csv'
outfile_json = 'D://data/michigan/mi-results-latest.jsonl'
mi_results.to_excel(outfile, index=False)
mi_results.to_csv(outfile_csv, index=False)
mi_results.to_json(outfile_json, orient='records', lines=True)
print('Saved Excel:', outfile)
print('Saved CSV:', outfile_csv)
print('Saved JSON:', outfile_json)

# Print out features.
features = {x: 'string' for x in mi_results.columns}
print('Number of features:', len(features))
print('Features:', features)


# === Analyze tests by month. ===

# Exclude outliers.
sample = mi_results.loc[
    (mi_results['total_thc'] > 0) &
    (mi_results['total_thc'] < 100) &
    (mi_results['product_type'] == 'Flower')
]
print('Number of samples:', len(sample))

# Visualize the frequency of tests by month/year.
test_frequency = sample['month'].value_counts().sort_index()
subsample = test_frequency[2:-1]
subsample.index = subsample.index.to_timestamp()
plt.figure(figsize=(12, 8))
sns.lineplot(
    x=subsample.index,
    y=subsample.values,
    marker="o",
    color="mediumblue"
)
plt.title('Monthly Number of Lab Tests in MI')
plt.ylabel('Number of Tests')
plt.xlabel('')
plt.xticks(rotation=45, ha='right')
plt.grid(True, which='both', linestyle='--', linewidth=0.5)
plt.tight_layout()
save_figure('mi-tests-by-month.png')
plt.show()


# === Analyze medical vs. adult-use testing. ===

# Visualize adult-use vs. medical tests over time.
grouped = sample.groupby(['month', 'medical']).size().reset_index(name='counts')
pivot_grouped = grouped.pivot(index='month', columns='medical', values='counts').fillna(0)
pivot_grouped = pivot_grouped.apply(pd.to_numeric, errors='coerce')
pivot_grouped.index = pivot_grouped.index.to_timestamp()
pivot_grouped = pivot_grouped[2:-1]
plt.figure(figsize=(15, 10))
for column in pivot_grouped.columns:
    sns.lineplot(data=pivot_grouped, x=pivot_grouped.index, y=column, marker='o', label=column)
plt.title('Number of Adult Use vs Medical Tests by Month in MI')
plt.ylabel('Number of Lab Tests')
plt.xlabel('')
plt.xticks(rotation=45, ha='right')
plt.legend(title='Adult Use / Medical')
plt.grid(True, which='both', linestyle='--', linewidth=0.5)
plt.tight_layout()
save_figure('mi-med-au-tests-by-month.png')
plt.show()

# Visualize the frequency distribution for medical.
subsample = sample[(sample['date'] >= datetime(2022, 1, 1)) &
                (sample['date'] < datetime(2023, 1, 1))]
med_au_distribution = subsample['medical'].value_counts()
plt.figure(figsize=(5, 8))
bar_plot = sns.barplot(x=med_au_distribution.index, y=med_au_distribution.values, palette='tab10')
plt.title('Adult-Use to Medical Lab Tests in MI in 2022', fontsize=21)
plt.ylabel('Number of Lab Tests')
plt.xlabel('')
for index, value in enumerate(med_au_distribution.values):
    bar_plot.text(index, value + 0.1, str(value), color='black', ha='center')
plt.tight_layout()
save_figure('mi-med-au-frequency.png')
plt.show()


# === Analyze lab market share. ===

# Count the number of labs.
labs = sample['lab'].unique()
print('Number of labs:', len(labs))

# Visualize the number of tests by lab.
subsample = sample[(sample['date'] >= datetime(2021, 1, 1)) &
                (sample['date'] < datetime(2022, 1, 1))]
lab_results = subsample.groupby('lab')
tests_by_lab = lab_results['total_thc'].count().sort_values(ascending=False)
sns.barplot(x=tests_by_lab.index, y=tests_by_lab.values, palette='tab20')
plt.xticks(rotation=45, ha='right')
plt.title('Lab Tests in MI in 2022')
plt.ylabel('Number of Lab Tests')
plt.xlabel('')
plt.tight_layout()
save_figure('mi-tests-by-lab.png')
plt.show()

# Visualize market share by lab in 2021.
subsample = sample[(sample['date'] >= datetime(2021, 1, 1)) &
                (sample['date'] < datetime(2022, 1, 1))]
lab_results = subsample.groupby('lab')
tests_by_lab = lab_results['total_thc'].count().sort_values(ascending=False)
market_share = tests_by_lab.div(tests_by_lab.sum()).mul(100).round(2)
sns.barplot(x=market_share.index, y=market_share.values, palette='tab20')
plt.title('Lab Market Share in MI in 2021')
plt.ylabel('Market Share (%)')
plt.xlabel('')
plt.xticks(rotation=45, ha='right')
plt.tight_layout()
save_figure('mi-market-share-by-lab-2021.png')
plt.show()

# Visualize market share by lab in 2022.
subsample = sample[(sample['date'] >= datetime(2022, 1, 1)) &
                (sample['date'] < datetime(2023, 1, 1))]
lab_results = subsample.groupby('lab')
tests_by_lab = lab_results['total_thc'].count().sort_values(ascending=False)
market_share = tests_by_lab.div(tests_by_lab.sum()).mul(100).round(2)
sns.barplot(x=market_share.index, y=market_share.values, palette='tab20')
plt.title('Lab Market Share in MI in 2022')
plt.ylabel('Market Share (%)')
plt.xlabel('')
plt.xticks(rotation=45, ha='right')
plt.tight_layout()
save_figure('mi-market-share-by-lab-2022.png')
plt.show()


# === Analyze total THC. ===

# Get a sub-sample.
subsample = sample[(sample['date'] >= datetime(2022, 1, 1)) &
                (sample['date'] < datetime(2023, 1, 1))]

# Visualize the distribution of THC.
mean_value = subsample['total_thc'].mean()
quantile_1 = subsample['total_thc'].quantile(0.01)
quantile_25 = subsample['total_thc'].quantile(0.25)
quantile_75 = subsample['total_thc'].quantile(0.75)
quantile_99 = subsample['total_thc'].quantile(0.99)
plt.figure(figsize=(12, 7))
sns.histplot(subsample['total_thc'], bins=100, color='lightblue', kde=True)
plt.axvline(quantile_1, color='blue', linestyle='dashed', linewidth=2, label=f'1st percentile: {quantile_1:.2f}%')
plt.axvline(quantile_25, color='green', linestyle='dashed', linewidth=2, label=f'25th percentile: {quantile_25:.2f}%')
plt.axvline(mean_value, color='red', linestyle='dashed', linewidth=2, label=f'Mean: {mean_value:.2f}%')
plt.axvline(quantile_75, color='darkgreen', linestyle='dashed', linewidth=2, label=f'75th percentile: {quantile_75:.2f}%')
plt.axvline(quantile_99, color='blue', linestyle='dashed', linewidth=2, label=f'99th percentile: {quantile_99:.2f}%')
plt.title('Total THC in MI Cannabis Flower in 2022', pad=15)
plt.xlabel('Total THC (%)')
plt.ylabel('Number of Tests')
plt.legend()
plt.tight_layout()
save_figure('mi-total-thc-distribution.png')
plt.show()

# Visualize the difference between medical and adult-use THC.
plt.figure(figsize=(12, 7))
sns.histplot(
    data=subsample,
    x='total_thc',
    hue='medical',
    bins=100,
    kde=True,
    palette={'Med': 'blue', 'AU': 'green'},
    stat='density',
)
median_med = subsample[subsample['medical'] == 'Med']['total_thc'].median()
median_au = subsample[subsample['medical'] == 'AU']['total_thc'].median()
plt.axvline(median_med, color='blue', linestyle='--', linewidth=1.5, label=f'Medical Median: {median_med:.2f}%')
plt.axvline(median_au, color='green', linestyle='--', linewidth=1.5, label=f'Adult-Use Median: {median_au:.2f}%')
plt.title('Total THC for Medical and Adult-Use in MI in 2022', pad=15)
plt.xlabel('Total THC (%)')
plt.ylabel('Frequency (%)')
plt.legend(loc='upper right', bbox_to_anchor=(1.3, 1))
plt.tight_layout()
save_figure('mi-med-au-total-thc-distribution.png')
plt.show()

# Perform a t-test to determine if the difference between medical and adult-use THC is significant.
med_thc = subsample[subsample['medical'] == 'Med']['total_thc']
au_thc = subsample[subsample['medical'] == 'AU']['total_thc']
t_stat, p_val = stats.ttest_ind(med_thc, au_thc, equal_var=True) # You can set equal_var to False for Welch's t-test
print(f'T-statistic: {t_stat}')
print(f'P-value: {p_val}')
alpha = 0.05
if p_val < alpha:
    print('The difference between medical and adult-use THC is statistically significant.')
else:
    print('The difference between medical and adult-use THC is not statistically significant.')


# === Analyze THC by lab. ===

# Visualize average THC percentage for each licensee.
average_thc_by_licensee = subsample.groupby('lab')['total_thc'].mean()
average_thc_by_licensee = average_thc_by_licensee.sort_values(ascending=False)
plt.figure(figsize=(25, 8))
bar_plot = sns.barplot(x=average_thc_by_licensee.index, y=average_thc_by_licensee.values, palette='tab20')
plt.title('Average Total THC by Lab in MI in 2022', pad=15)
plt.ylabel('Average Total THC (%)')
plt.xlabel('Lab')
plt.xticks(rotation=45, ha='right')
for index, value in enumerate(average_thc_by_licensee.values):
    bar_plot.text(index, value + 0.2, f'{value:.0f}%', color='black', ha='center')
mean = average_thc_by_licensee.mean()
plt.axhline(
    y=mean,
    color='red',
    linestyle='--',
    label=f'MI Avg Total THC: {mean:.2f}%',
)
plt.tight_layout()
save_figure('mi-total-thc-by-lab.png')
plt.show()


# === Augment strain data. ===

def extract_strain_name(product_name):
    """Extract the strain name from the product name."""
    name = str(product_name)
    strain_name = re.split(r' - | \| | _ | x | – | — |:|\(|\)|/', name)[0]
    strain_name = strain_name.split('Buds')[0].strip()
    strain_name = strain_name.split('Bulk')[0].strip()
    strain_name = strain_name.split('Flower')[0].strip()
    strain_name = strain_name.split('Pre-Roll')[0].strip()
    return strain_name


# Augment strain names.
sample['strain_name'] = sample['product_name'].apply(extract_strain_name)
print(sample.sample(10)['strain_name'])


# === Analyze strains. ===

# Exclude samples with strain_name set to None, '' or 'Unprocessed'
sample = sample[sample['strain_name'].notna()]
sample = sample[~sample['strain_name'].isin(['', 'Unprocessed'])]

# Standardize strain names
sample['strain_name'] = sample['strain_name'].replace({
    'Gorilla Glue': 'Gorilla Glue #4',
    'GG4': 'Gorilla Glue #4'
})

# Restrict the timeframe to 2022.
subsample = sample[(sample['date'] >= datetime(2022, 1, 1)) &
                (sample['date'] < datetime(2023, 1, 1))]

# Visualize the frequency of each strain
strain_counts = subsample['strain_name'].value_counts()
counts = strain_counts.head(20)
plt.figure(figsize=(13, 13))
bar_plot = sns.barplot(
    y=counts.index,
    x=counts.values,
    palette='tab20',
)
plt.title('Number of Lab Tests for the Top 20 Strains in MI in 2022', pad=15)
plt.xlabel('')
plt.ylabel('')
for index, value in enumerate(counts.values):
    bar_plot.text(value, index, str(value), color='black', ha='left', va='center')
plt.tight_layout()
save_figure('mi-top-strains.png')
plt.show()

# Visualize the average THC for the top strains.
avg_thc_per_strain = subsample.groupby('strain_name')['total_thc'].mean().sort_values(ascending=False)
overall_avg_thc = subsample['total_thc'].mean()
print('Overall average THC:', round(overall_avg_thc, 2))
print('99th percentile THC:', round(sample['total_thc'].quantile(0.99), 2))
top_20_strains = strain_counts.head(20).index
avg_thc_top_20_strains = avg_thc_per_strain[avg_thc_per_strain.index.isin(top_20_strains)]
avg_thc_top_20_strains = avg_thc_top_20_strains.loc[top_20_strains]
print('Average THC for top 20 strains:', round(avg_thc_top_20_strains.mean(), 2))
plt.figure(figsize=(26, 10))
bar_plot = sns.barplot(
    x=avg_thc_top_20_strains.index,
    y=avg_thc_top_20_strains.values,
    palette='tab20'
)
plt.axhline(
    y=overall_avg_thc,
    color='red',
    linestyle='--',
    label=f'MI Avg Total THC: {overall_avg_thc:.2f}%',
)
plt.title('Average Total THC for the Top 20 Strains in MI in 2022', fontsize=36, pad=15)
plt.ylabel('Total THC (%)')
plt.xlabel('')
plt.xticks(rotation=45, ha='right')
plt.legend()
for p in bar_plot.patches:
    bar_plot.annotate(format(p.get_height(), '.2f') + '%', 
                      (p.get_x() + p.get_width() / 2., p.get_height()), 
                      ha='center', va='center', 
                      xytext=(0, 9), 
                      textcoords='offset points')
plt.tight_layout()
save_figure('mi-avg-thc-by-top-20-strains.png')
plt.show()

# Look at the top adult-use strains.
adult_use = subsample.loc[subsample['medical'] == 'AU']
strain_counts = adult_use['strain_name'].value_counts()
avg_thc_per_strain = adult_use.groupby('strain_name')['total_thc'].mean().sort_values(ascending=False)
overall_avg_thc = adult_use['total_thc'].mean()
print('Adult-use average THC:', round(overall_avg_thc, 2))
print('Adult-use 99th percentile THC:', round(sample['total_thc'].quantile(0.99), 2))
top_20_strains = strain_counts.head(20).index
avg_thc_top_20_strains = avg_thc_per_strain[avg_thc_per_strain.index.isin(top_20_strains)]
avg_thc_top_20_strains = avg_thc_top_20_strains.loc[top_20_strains]
print('Average THC for top 20 adult-use strains:', round(avg_thc_top_20_strains.mean(), 2))
plt.figure(figsize=(26, 10))
bar_plot = sns.barplot(
    x=avg_thc_top_20_strains.index,
    y=avg_thc_top_20_strains.values,
    palette='tab20'
)
plt.axhline(
    y=overall_avg_thc,
    color='red',
    linestyle='--',
    label=f'MI Adult-Use Avg Total THC: {overall_avg_thc:.2f}%',
)
plt.title('Average Total THC for the Top 20 Adult-Use Strains in MI in 2022', fontsize=36, pad=15)
plt.ylabel('Total THC (%)')
plt.xlabel('')
plt.xticks(rotation=45, ha='right')
plt.legend()
for p in bar_plot.patches:
    bar_plot.annotate(format(p.get_height(), '.2f') + '%', 
                      (p.get_x() + p.get_width() / 2., p.get_height()), 
                      ha='center', va='center', 
                      xytext=(0, 9), 
                      textcoords='offset points')
plt.tight_layout()
save_figure('mi-avg-thc-by-top-20-strains-adult-use.png')
plt.show()

# Look at the top medical strains.
medical = subsample.loc[subsample['medical'] == 'Med']
strain_counts = medical['strain_name'].value_counts()
avg_thc_per_strain = medical.groupby('strain_name')['total_thc'].mean().sort_values(ascending=False)
overall_avg_thc = medical['total_thc'].mean()
print('Medical average THC:', round(overall_avg_thc, 2))
print('Medical 99th percentile THC:', round(sample['total_thc'].quantile(0.99), 2))
top_20_strains = strain_counts.head(20).index
avg_thc_top_20_strains = avg_thc_per_strain[avg_thc_per_strain.index.isin(top_20_strains)]
avg_thc_top_20_strains = avg_thc_top_20_strains.loc[top_20_strains]
print('Average THC for top 20 medical strains:', round(avg_thc_top_20_strains.mean(), 2))
plt.figure(figsize=(26, 10))
bar_plot = sns.barplot(
    x=avg_thc_top_20_strains.index,
    y=avg_thc_top_20_strains.values,
    palette='tab20'
)
plt.axhline(
    y=overall_avg_thc,
    color='red',
    linestyle='--',
    label=f'MI Medical Avg Total THC: {overall_avg_thc:.2f}%',
)
plt.title('Average Total THC for the Top 20 Medical Strains in MI in 2022', fontsize=36, pad=15)
plt.ylabel('Total THC (%)')
plt.xlabel('')
plt.xticks(rotation=45, ha='right')
plt.legend()
for p in bar_plot.patches:
    bar_plot.annotate(format(p.get_height(), '.2f') + '%', 
                      (p.get_x() + p.get_width() / 2., p.get_height()), 
                      ha='center', va='center', 
                      xytext=(0, 9), 
                      textcoords='offset points')
plt.tight_layout()
save_figure('mi-avg-thc-by-top-20-strains-med.png')
plt.show()