app.py

import streamlit as st
import pandas as pd
import numpy as np

def composite_correlations(R, composite_idx, var_names=None, augment=False):
    """Compute unit-weighted composite correlations."""
    R_mat = np.asarray(R, dtype=float)
    n_all = R_mat.shape[0]
    n = len(composite_idx)
    # sub-matrix of the composite items
    R_yy = R_mat[np.ix_(composite_idx, composite_idx)]
    # mean off-diagonal
    iu = np.triu_indices(n, k=1)
    rbar = R_yy[iu].mean() if iu[0].size > 0 else 0.0
    denom = np.sqrt(n + n*(n-1)*rbar)
    numer = R_mat[composite_idx, :].sum(axis=0)
    r_comp = numer / denom
    if var_names is not None:
        r_comp = pd.Series(r_comp, index=var_names, name="Composite")
    if not augment:
        return r_comp
    # build augmented matrix
    if var_names is not None:
        idx = list(var_names) + ["Composite"]
        R_aug = pd.DataFrame(np.zeros((n_all+1, n_all+1)), index=idx, columns=idx)
        R_aug.iloc[:-1, :-1] = R_mat
        R_aug.iloc[-1, :-1] = r_comp.values
        R_aug.iloc[:-1, -1] = r_comp.values
        R_aug.iloc[-1, -1] = 1.0
    else:
        R_aug = np.zeros((n_all+1, n_all+1))
        R_aug[:n_all, :n_all] = R_mat
        R_aug[n_all, :n_all] = r_comp
        R_aug[:n_all, n_all] = r_comp
        R_aug[n_all, n_all] = 1.0
    return r_comp, R_aug

import os
# On Huggingface Spaces the home directory may be unwritable; override it to the current working directory
os.environ['HOME'] = os.getcwd()
# Disable streamlit usage stats to avoid write attempts
os.environ['STREAMLIT_BROWSER_GATHER_USAGE_STATS'] = 'false'

# Streamlit UI
st.title("Composite-Correlation Calculator")
st.markdown(
    """
    Upload a CSV containing your (possibly lower-triangular) correlation matrix.  
    The app will fill in missing cells by symmetry, set diagonals to 1,  
    then let you select two sets of variables to form two unit-weighted composites.
    """
)

uploaded = st.file_uploader("Upload correlation matrix (CSV)", type=["csv"])
if uploaded is not None:
    # 1) read and label
    try:
        df = pd.read_csv(uploaded, index_col=0)
    except Exception:
        st.error("Failed to read CSV. Make sure the first column contains row labels.")
        st.stop()
    if df.shape[0] != df.shape[1]:
        st.error("Matrix must be square.")
        st.stop()

    st.success(f"Loaded a {df.shape[0]}×{df.shape[1]} matrix.")

    # 2) symmetrize and fill diagonal
    mat = df.values.astype(float)
    mat = np.where(np.isnan(mat), mat.T, mat)
    np.fill_diagonal(mat, 1.0)
    df_sym = pd.DataFrame(mat, index=df.index, columns=df.columns)
    st.write("**Symmetrized & filled diagonal:**")
    st.dataframe(df_sym)

    all_vars = list(df_sym.columns)
    cols1, cols2 = st.columns(2)
    with cols1:
        group1 = st.multiselect(
            "Select variables for Composite 1", 
            options=all_vars,
            default=all_vars[: min(3, len(all_vars))],
            key='g1'
        )
    with cols2:
        group2 = st.multiselect(
            "Select variables for Composite 2", 
            options=all_vars,
            default=all_vars[: min(3, len(all_vars))],
            key='g2'
        )

    ok1 = len(group1) >= 2
    ok2 = len(group2) >= 2
    if not ok1 or not ok2:
        st.warning("Pick at least 2 variables for each composite.")
    else:
        if st.button("Compute composites and their correlations"):
            idx1 = [all_vars.index(v) for v in group1]
            idx2 = [all_vars.index(v) for v in group2]
            # compute each composite vs all vars
            r_comp1 = composite_correlations(
                df_sym.values, composite_idx=idx1, var_names=all_vars, augment=False
            )
            r_comp2 = composite_correlations(
                df_sym.values, composite_idx=idx2, var_names=all_vars, augment=False
            )
            # compute composite vs composite
            # numerator: sum R[i,j] i in idx1, j in idx2
            R_mat = df_sym.values
            numer_cc = R_mat[np.ix_(idx1, idx2)].sum()
            # denom: group1 denom and group2 denom
            # reuse denom calculation
            def denom_for(idx):
                sub = R_mat[np.ix_(idx, idx)]
                iu = np.triu_indices(len(idx), k=1)
                rbar = sub[iu].mean() if iu[0].size>0 else 0.0
                return np.sqrt(len(idx) + len(idx)*(len(idx)-1)*rbar)
            denom1 = denom_for(idx1)
            denom2 = denom_for(idx2)
            r_cc = numer_cc/(denom1*denom2)

            st.subheader("Composite 1 vs. Each Variable")
            st.dataframe(r_comp1.to_frame(name="Comp1"))
            st.subheader("Composite 2 vs. Each Variable")
            st.dataframe(r_comp2.to_frame(name="Comp2"))
            st.subheader("Composite1 vs Composite2 Correlation")
            st.write(f"**r = {r_cc:.4f}**")

            st.subheader("Augmented Correlation Matrix (with Composites)")
            # build augmented matrix with two composites
            idx = all_vars + ["Comp1", "Comp2"]
            R_aug2 = pd.DataFrame(
                np.zeros((len(all_vars)+2, len(all_vars)+2)), index=idx, columns=idx
            )
            R_aug2.iloc[:len(all_vars), :len(all_vars)] = R_mat
            R_aug2.loc["Comp1", all_vars] = r_comp1.values
            R_aug2.loc[all_vars, "Comp1"] = r_comp1.values
            R_aug2.loc["Comp1", "Comp1"] = 1.0
            R_aug2.loc["Comp2", all_vars] = r_comp2.values
            R_aug2.loc[all_vars, "Comp2"] = r_comp2.values
            R_aug2.loc["Comp2", "Comp2"] = 1.0
            R_aug2.loc["Comp1", "Comp2"] = r_cc
            R_aug2.loc["Comp2", "Comp1"] = r_cc
            st.dataframe(R_aug2)
else:
    st.info("🤖 Upload a CSV file to get started.")