pams-kim 0.1.0

Profile Analysis via Multidimensional Scaling

pams-kim

Profile Analysis via Multidimensional Scaling (PAMS)

Python port of the R pams package (CRAN, Kim & Kim 2026).

---

Overview

PAMS identifies population-level core response profiles from cross-sectional or longitudinal person-score data using nonmetric multidimensional scaling (SMACOF algorithm). Each person profile is decomposed into:

• Level component: person mean across subscales
• Pattern component: ipsatized subscores (within-person variation)

PAMS fits a nonmetric MDS solution to the J x J inter-variable distance matrix, bootstraps the solution to generate empirical sampling distributions, and computes BCa confidence intervals for each core profile coordinate.

Key features:

• Nonmetric MDS via SMACOF (de Leeuw & Mair, 2009) or classical metric MDS
• BCa bootstrap CIs for core profile coordinates
• Person-level weights, R-squared, and profile correlations
• Optional individual bootstrap CIs for selected participants
• Longitudinal support (time-ordered subscale columns)

References: Kim & Kim (2024) doi:10.20982/tqmp.20.3.p230

---

Installation

pip install pams-kim

Requirements: Python >= 3.9, numpy >= 1.24, scipy >= 1.10, scikit-learn >= 1.3

---

Quick start

import numpy as np
from pams import BootSmacof

# Toy example (matches R PAMS documentation)
rng = np.random.default_rng(42)
toy_data = rng.normal(10, 2, (50, 5))

result = BootSmacof(
    testdata    = toy_data,
    participant = [0, 1, 2],   # row indices for individual CIs
    mds         = "smacof",
    type        = "ordinal",
    distance    = "euclid",
    nprofile    = 2,
    direction   = [1, 1],
    cl          = 0.95,
    nBoot       = 2000,
    testname    = ["S1", "S2", "S3", "S4", "S5"],
)

print(f"Stress: {result['MDS']['stress']:.4f}")
print(f"Mean R2: {result['WeightmeanR2']:.3f}")
print(result["MDSsummary"][0].round(3))  # Core Profile 1 with BCa CIs

---

Output fields

result["MDS"] # dict with conf (J x K) and stress result["MDSsummary"] # list of K DataFrames -- BCa CIs per core profile result["MDSprofile"] # list of K arrays (nBoot, J) -- bootstrap distributions result["stresssummary"] # one-row DataFrame of bootstrap stress result["MDSR2"] # (K,) R² confirming non-collinearity of core profiles result["Weight"] # (n, 2K+2) person weights, level, R², correlations result["WeightmeanR2"] # mean person R² across all persons result["WeightB"] # bootstrap weight summary for participants result["PcorrB"] # bootstrap partial correlations for participants

---

Lower-level functions

from pams import (
    intervar_distance,  # J x J inter-variable distance matrix
    smacof_mds,         # SMACOF nonmetric MDS
    classical_mds,      # Torgerson classical metric MDS
    apply_direction,    # sign flipping for core profiles
)

---

R package

CRAN: https://cran.r-project.org/package=pams GitHub: https://github.com/sekangakim/pams

---

Citation

Kim, S.-K., & Kim, D. (2024). Utility of profile analysis via multidimensional scaling in R for the study of person response profiles in cross-sectional and longitudinal data. The Quantitative Methods for Psychology, 20(3), 230-247. doi:10.20982/tqmp.20.3.p230

---

License

MIT

Authors

Se-Kang Kim, Ph.D. -- [e-mail] Baylor College of Medicine / Texas Children's Hospital

Donghoh Kim, Ph.D.