# SHAP Analysis Reveals Persistent Gender Orgasm Gap Across Lifespan

> Published 2026-10-23 · https://www.promptzone.com/dalia_delgado/shap-analysis-reveals-persistent-gender-orgasm-gap-across-lifespan-4o41

A machine learning analysis of sexual satisfaction data, [flagged on Hacker News](https://pmc.ncbi.nlm.nih.gov/articles/PMC11217024/), quantifies the persistent orgasm gap between men and women across adult life. Published on April 29, 2024, in *PLOS ONE*, the study of **2,213 Finnish adults** aged 18-78 applies SHAP (SHapley Additive exPlanations) to identify the most impactful predictors of orgasm frequency.

> **Study:** The Lifelong Orgasm Gap | **Sample:** 2,213 adults | **Method:** XGBoost + SHAP
> **Key Metric:** Orgasm Frequency (1-7 scale) | **License:** CC BY 4.0

## How Machine Learning Mapped Intimate Data
Researchers used an **XGBoost regression model** to analyze survey data from the Finnish Genetic Sexual Dysfunction Study. The model predicted self-reported orgasm frequency based on **87 variables** covering demographics, health, sexual behavior, and psychological factors.

The critical analytical step was applying **SHAP value analysis** post-modeling. This game theory-based method quantifies the contribution of each feature to the model's predictions for individual participants. Unlike simple correlation, SHAP reveals **non-linear relationships** and interaction effects, showing *how* and *how much* a factor like age or communication influences the outcome.

> **Bottom line:** SHAP analysis transformed survey data into an interpretable map of what factors drive orgasm frequency, and for whom.


![SHAP Analysis Reveals Persistent Gender Orgasm Gap Across Lifespan](https://promptzone-community.s3.amazonaws.com/uploads/articles/q7w98fhpquf0vg95erhs.jpeg)

## The Numbers: Gender Gap Persists at All Ages
The data confirms a significant average gap. Men reported a mean orgasm frequency of **5.7** on the 1-7 scale, compared to **4.5** for women. The SHAP analysis revealed this gender difference was one of the strongest overall predictors.

Crucially, the gap did not disappear with age. While life satisfaction and relationship length increased in importance for older cohorts, gender remained a dominant factor. For women, factors like **sexual communication** and the **importance of orgasm** to one's partner had higher positive SHAP values, indicating a greater relative impact on their reported frequency.

| Predictor | High Positive Impact (SHAP) For | High Negative Impact (SHAP) For |
|-----------|--------------------------------|---------------------------------|
| Gender (Male) | Men's Scores | Women's Scores |
| Sexual Communication | Women | – |
| Partner's Orgasm Importance | Women | – |
| Psychological Distress | – | All |
| Masturbation Frequency | Men | – |

## How to Explore the Methodology and Data
For practitioners interested in replicating or extending this analytical approach, the full resources are publicly available. This provides a concrete dataset for practicing explainable AI (XAI) techniques on complex behavioral data.

{% details "Access the study materials" %}
- **Full Paper & Data:** Available on [PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC11217024/) under a CC BY 4.0 license.
- **Analysis Code:** The R code for the XGBoost and SHAP analysis is included in the paper's Supporting Information files (S1 File).
- **SHAP Libraries:** To implement similar analysis, use the [`shap` Python library](https://github.com/shap/shap) or the `SHAPforxgboost` package in R.
{% enddetails %}

## Pros and Cons of the AI Approach
This study demonstrates both the power and the limits of applying modern ML tools to sensitive sociological data.

**Strengths:**
- Moves beyond simple statistics to **model complex interactions** between dozens of variables.
- **SHAP provides individual-level explanations**, highlighting that predictors like "sexual communication" don't affect all women equally.
- The open data and code provide a **benchmark dataset** for XAI method development.

**Limitations:**
- **Correlation, not causation:** The model identifies associations but cannot prove what directly causes higher orgasm frequency.
- **Cultural specificity:** Data is from Finland; predictors may differ in other cultural contexts.
- **Self-reported data** carries inherent biases in recall and social desirability.

## Alternatives and Comparison to Other Analytical Methods
This study's choice of XGBoost + SHAP is one of several viable approaches for explainable analysis of tabular data. The methodology selection involves clear trade-offs.

| Method | Best For | Explainability | Use Case vs. This Study |
|--------|----------|----------------|--------------------------|
| **XGBoost + SHAP** | High-accuracy tabular data, non-linear relationships | Post-hoc, local & global | **Used here:** Balances performance with detailed feature attribution. |
| **Logistic Regression** | Simpler models, hypothesis testing | Intrinsic (coefficients) | Less accurate for complex interactions but easier to interpret causally. |
| **Deep Neural Net + LIME** | Very high-dimensional data | Post-hoc, local approximations | Overkill for 87 features; LIME is less consistent than SHAP. |
| **Standard Statistical Tests (ANOVA)** | Comparing group means | Descriptive only | Would miss the interactive effects captured by ML. |

## Who Should Use This Research Methodology
This paper is a practical case study for specific AI practitioner audiences, but less relevant for others.

**Highly relevant for:**
- **Data scientists** building explainable models for behavioral or healthcare tabular data.
- **Researchers** in sociology or public health looking to adopt modern ML explainability tools.
- **Educators** seeking a real-world, accessible dataset to teach SHAP and XGBoost.

**Can likely skip if:**
- You work exclusively with unstructured data (NLP, CV).
- Your focus is on deploying high-throughput predictive models where explainability is secondary.
- You require causal inference methodologies like randomized controlled trials.

## Bottom Line: A Blueprint for Explainable Social Science
The study's primary contribution for the AI community is not its sociological finding, but its **methodological demonstration**. It successfully adapts tools from the ML engineering toolkit—XGBoost and SHAP—to a messy, real-world social science problem, producing interpretations that are more nuanced than traditional statistics. It showcases how to move from predicting an outcome to understanding the model's decision-making process across a population.

The open dataset serves as a valuable bridge, inviting technical practitioners to test other XAI methods on a grounded, human-centric problem. The persistent gap it quantifies is a societal issue, but the analytical path it charts offers a template for future research.