Skip to content

Commit

Permalink
Merge pull request #28 from Bayer-Group/streamlit
Browse files Browse the repository at this point in the history 
Streamlit
  • Loading branch information
@sprivite
sprivite authored Jul 8, 2024
2 parents 0fb3ce1 + 46639f5 commit 0f106ee
Show file tree
Hide file tree
Showing 2 changed files with 233 additions and 115 deletions.
346 changes: 232 additions & 114 deletions bin/main.py
View file
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
import streamlit as st
import pyarrow as pa
import pandas as pd
import seaborn as sns

from pybalance.propensity import PropensityScoreMatcher
Expand All @@ -9,63 +9,63 @@
plot_categoric_features,
plot_per_feature_loss,
)
from pybalance.utils import BALANCE_CALCULATORS
from pybalance.utils import BALANCE_CALCULATORS, split_target_pool, MatchingData

OBJECTIVES = list(BALANCE_CALCULATORS.keys())
OBJECTIVES.remove("base")

# Create a Streamlit app
# st.title('Propensity Score Matcher App')
col1, col2, col3 = st.columns([1, 3, 1])
with col2:
st.image("logo.png", width=400)
st.markdown(
"<h5> <center>The matching toolkit for observational data.</center></h5>",
unsafe_allow_html=True,
)
placeholder = st.empty()


# Create a sidebar for inputting parameters
st.sidebar.title("Parameters")
n_pool = st.sidebar.number_input(
"Pool size",
min_value=1,
step=1000,
value=10000,
help="Number of patients in the pool population",
)
n_target = st.sidebar.number_input(
"Target size",
min_value=1,
step=100,
value=1000,
help="Number of patients in the target population",
)
# seed = st.sidebar.number_input('Random Seed', min_value=0, step=1, value=45, help='Random seed for dataset generation')
seed = 45
objective = st.sidebar.selectbox("Objective", OBJECTIVES)
# caliper = st.sidebar.number_input('Caliper', min_value=0.0, max_value=1.0, value=1.0, step=0.01, help='If defined, restricts matches to those patients with propensity scores within the caliper of each other')
max_iter = st.sidebar.number_input(
"Max Iterations",
min_value=1,
help="Maximum number of hyperparameters to try before returning the best match",
)
time_limit = st.sidebar.number_input(
"Time Limit",
min_value=10,
help="Restrict hyperparameter search based on time in seconds",
)
method = st.sidebar.selectbox("Method", ["greedy", "linear_sum_assignment"])
cumulative = st.sidebar.checkbox("Cumulative plots", value=False)
if cumulative:
bins = 500
else:
bins = 10
def generate_data():
print("Generating data!")
seed = 45
n_pool, n_target = st.session_state["n_pool"], st.session_state["n_target"]
matching_data = generate_toy_dataset(n_pool, n_target, seed)
st.session_state["matching_data"] = matching_data
st.session_state["first_run"] = False


def load_data():
uploaded_file = st.session_state.get("uploaded_file")
if uploaded_file is None:
st.warning(f"Please specify a file to upload.")
return
population_column = st.session_state.get("population_column")
try:
df = pd.read_csv(uploaded_file)
except: # FIXME catch specific exception UnicodeDecodeError
df = pd.read_parquet(uploaded_file)
matching_data = MatchingData(
df,
population_col=population_column,
)
try:
split_target_pool(matching_data)
except:
st.warning(
f"Cannot split data based on {population_column}. Please specify a split column with two unique values."
)
return

palette = sns.color_palette("colorblind")
st.session_state["first_run"] = False
st.session_state["matching_data"] = matching_data

# Update the parameters based on user input
matching_data = generate_toy_dataset(n_pool, n_target, seed)
pre_matching_data = matching_data.copy()
hue_order = list(matching_data.populations)

# Create a button to trigger the match() method
if st.sidebar.button("Match"):
def match():

# Create an instance of PropensityScoreMatcher
max_iter = st.session_state.get("max_iter", 100)
method = "greedy"
objective = st.session_state.get("objective")
matching_data = st.session_state.get("matching_data").copy()
matcher = PropensityScoreMatcher(
matching_data, objective, None, max_iter, time_limit, method
)
Expand All @@ -75,73 +75,191 @@
post_matching_data.data.loc[:, "population"] = (
post_matching_data["population"] + " (postmatch)"
)
matching_data.append(post_matching_data.data)

balance_calculator = BALANCE_CALCULATORS[objective](pre_matching_data)
st.sidebar.write(balance_calculator.__doc__)
hue_order += list(set(matching_data.populations) - set(hue_order))

# Display the figures
if matching_data:

tab1, tab2, tab3 = st.tabs(["Numeric", "Categoric", "SMD"])
with tab1:

plot_vars = []
for i, col in enumerate(st.columns(len(matching_data.headers["numeric"]))):
with col:
col_name = matching_data.headers["numeric"][i]
if st.checkbox(col_name, value=True):
plot_vars.append(col_name)
print("streamlit", plot_vars)
numeric_fig = plot_numeric_features(
matching_data,
col_wrap=2,
height=6,
hue_order=hue_order,
cumulative=cumulative,
bins=bins,
include_only=plot_vars,
# palette=palette,
st.session_state["post_matching_data"] = post_matching_data


def load_front_page():

st.markdown("<h5>Generate a simulated dataset</h5>", unsafe_allow_html=True)

col1, col2 = st.columns(2)
with col1:
st.number_input(
"Pool size",
min_value=1,
step=1000,
value=10000,
key="n_pool",
help="Number of patients in the pool (by convention, larger) population",
)
with col2:
st.number_input(
"Target size",
min_value=1,
step=100,
value=1000,
key="n_target",
help="Number of patients in the target (by convention, smaller) population",
)
st.button("Generate", on_click=generate_data)

st.write("---")
st.markdown("<h5>Upload your own data</h5>", unsafe_allow_html=True)
col1, col2 = st.columns([3, 1])
with col1:
st.file_uploader(
"Upload file",
type=["csv", "parquet"],
help="Choose a matching file to upload.",
key="uploaded_file",
)
with col2:
st.text_input(
"population column",
value="population",
key="population_column",
help="Column name for splitting data into different populations",
)
st.button("Submit", on_click=load_data)
st.write(
"We retain your data only for the duration of your user session. You are responsible for the security of your data and compliance with your data license."
)


# Load landing page
if st.session_state.get("first_run", True):
with placeholder.container():
load_front_page()

if not st.session_state.get("first_run", True):
with placeholder.container():

matching_data = st.session_state.get("matching_data").copy()
target, pool = split_target_pool(matching_data)

# Create a sidebar for inputting parameters
st.sidebar.title("Matching parameters")
st.sidebar.number_input(
"Matched pool size",
min_value=1,
step=100,
value=len(target),
max_value=len(pool),
key="n_pool_matched",
help="Number of patients in the pool population after matching",
)
st.sidebar.number_input(
"Matched Target size",
min_value=1,
step=100,
value=len(target),
max_value=len(target),
key="n_target_matched",
help="Number of patients in the target population after matching",
)
st.pyplot(numeric_fig)
st.write("---")
# import pdb
# pdb.set_trace()
summary = matching_data.describe_numeric().astype("object")
summary = summary[summary.index.get_level_values(0).isin(plot_vars)]
st.dataframe(summary, use_container_width=True)

with tab2:
plot_vars = []
for i, col in enumerate(st.columns(len(matching_data.headers["categoric"]))):
with col:
col_name = matching_data.headers["categoric"][i]
if st.checkbox(col_name, value=True):
plot_vars.append(col_name)

print("streamlit", plot_vars)
categoric_fig = plot_categoric_features(
matching_data,
col_wrap=2,
height=6,
include_binary=True,
hue_order=hue_order,
include_only=plot_vars,
# palette=palette,
# seed = st.sidebar.number_input('Random Seed', min_value=0, step=1, value=45, help='Random seed for dataset generation')
objective = st.sidebar.selectbox("Objective", OBJECTIVES, key="objective")
# caliper = st.sidebar.number_input('Caliper', min_value=0.0, max_value=1.0, value=1.0, step=0.01, help='If defined, restricts matches to those patients with propensity scores within the caliper of each other')
st.sidebar.number_input(
"Max Iterations",
min_value=1,
key="max_iter",
help="Maximum number of hyperparameters to try before returning the best match",
)
st.pyplot(categoric_fig)
st.write("---")
summary = matching_data.describe_categoric().astype("object")
summary = summary[summary.index.get_level_values(0).isin(plot_vars)]
st.dataframe(summary, use_container_width=True)

with tab3:
categoric_fig = plot_per_feature_loss(
matching_data,
balance_calculator,
hue_order=hue_order,
debin=False,
# palette=palette,
time_limit = st.sidebar.number_input(
"Time Limit",
min_value=10,
key="time_limit",
help="Restrict hyperparameter search based on time in seconds",
)
st.pyplot(categoric_fig)
# method = st.sidebar.selectbox("Method", ["greedy", "linear_sum_assignment"])
cumulative = st.sidebar.checkbox("Cumulative plots", value=False)
if cumulative:
bins = 500
else:
bins = 10

palette = sns.color_palette("colorblind")

# Update the parameters based on user input
hue_order = list(matching_data.populations)

# Create a button to trigger the match() method
st.sidebar.button("Match", on_click=match)

balance_calculator = BALANCE_CALCULATORS[objective](matching_data)
st.sidebar.write(balance_calculator.__doc__)

if "post_matching_data" in st.session_state.keys():
post_matching_data = st.session_state["post_matching_data"]
matching_data.append(post_matching_data.data)

hue_order += list(set(matching_data.populations) - set(hue_order))
print("populations", hue_order)
tab1, tab2, tab3, tab4 = st.tabs(["Numeric", "Categoric", "SMD", "Summary"])
with tab1:

plot_vars = []
for i, col in enumerate(st.columns(len(matching_data.headers["numeric"]))):
with col:
col_name = matching_data.headers["numeric"][i]
if st.checkbox(col_name, value=True):
plot_vars.append(col_name)
print("streamlit", plot_vars)
numeric_fig = plot_numeric_features(
matching_data,
col_wrap=2,
height=6,
hue_order=hue_order,
cumulative=cumulative,
bins=bins,
include_only=plot_vars,
# palette=palette,
)
st.pyplot(numeric_fig)
st.write("---")
# import pdb
# pdb.set_trace()
summary = matching_data.describe_numeric().astype("object")
summary = summary[summary.index.get_level_values(0).isin(plot_vars)]
st.dataframe(summary, use_container_width=True)

with tab2:
plot_vars = []
for i, col in enumerate(
st.columns(len(matching_data.headers["categoric"]))
):
with col:
col_name = matching_data.headers["categoric"][i]
if st.checkbox(col_name, value=True):
plot_vars.append(col_name)

print("streamlit", plot_vars)
categoric_fig = plot_categoric_features(
matching_data,
col_wrap=2,
height=6,
include_binary=True,
hue_order=hue_order,
include_only=plot_vars,
# palette=palette,
)
st.pyplot(categoric_fig)
st.write("---")
summary = matching_data.describe_categoric().astype("object")
summary = summary[summary.index.get_level_values(0).isin(plot_vars)]
st.dataframe(summary, use_container_width=True)

with tab3:
categoric_fig = plot_per_feature_loss(
matching_data,
balance_calculator,
hue_order=hue_order,
debin=False,
# palette=palette,
)
st.pyplot(categoric_fig)

with tab4:
st.markdown(matching_data._repr_html_(), unsafe_allow_html=True)
st.dataframe(matching_data.counts(), use_container_width=True)
2 changes: 1 addition & 1 deletion pybalance/utils/balance_calculators.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -317,7 +317,7 @@ def per_feature_loss(

class BetaBalance(BaseBalanceCalculator):
"""
BetaBalance computes the balance between two populatiosn as the mean
BetaBalance computes the balance between two populations as the mean
absolute standardized mean difference across all features. Uses
StandardMatchingPreprocessor as the preprocessor. In this preprocessor,
numeric variables are left unchanged, while categorical variables are
Expand Down

0 comments on commit 0f106ee

Please sign in to comment.