November 2026

Research Question

Does the amount of sleep a person gets relate to their reported depressive symptoms?

Main research question:

Is average sleep duration associated with reported depressive symptoms?

This analysis examines weekday and weekend sleep patterns using data from the CDC’s National Health and Nutrition Examination Survey (NHANES).

Important: This study examines associations, not cause and effect.

Data Source and Variables

Data source: CDC National Health and Nutrition Examination Survey (NHANES), 2017–2018.

I combined two data sets using each participant’s unique identification number (SEQN).

Variables examined:

  • SLD012: Usual weekday/workday sleep duration (hours).
  • SLD013: Usual weekend/non-workday sleep duration (hours).
  • DPQ010–DPQ090: Nine questions measuring depressive symptoms.

Sample size: 5,032 participants with valid weekday sleep and depressive symptom measurements.

I also examined weekend sleep, which reduced the sample to 5,015 participants.

Measuring Depressive Symptoms

The Patient Health Questionnaire (PHQ-9) is a screening tool used to measure depressive symptoms.

Each question asks how frequently a symptom occurred over the previous two weeks:

  • 0 = Not at all
  • 1 = Several days
  • 2 = More than half the days
  • 3 = Nearly every day

For this analysis, eight of the nine questions were combined to create a modified depressive symptom score.

Why was one question excluded?

The sleep-related question (DPQ030) was excluded to avoid directly including sleep in both the predictor and outcome measurements.

Higher scores indicate more frequently reported depressive symptoms.

Sleep Duration and Depression Scores

An investigation into whether nightly sleep duration is associated with reported depressive symptoms.

## `geom_smooth()` using formula = 'y ~ x'

## Distribution of Depression Scores {.smaller}

Interactive 3D Sleep and Depression Plot

Linear Regression

Linear regression was used to examine whether sleep duration could help explain differences in depressive symptom scores.

The model estimates the relationship between sleep duration (X) and depressive symptom score (Y).

Regression equation:

\[ Y = \beta_0 + \beta_1X + \epsilon \]

  • Y = Predicted depressive symptom score
  • X = Sleep duration in hours
  • β₀ = Intercept
  • β₁ = Change in score associated with one additional hour of sleep
  • ε = Unexplained variation

A p-value below 0.05 was used as the threshold for statistical significance.

R Code: Linear Regression

The lm() function was used to create a linear regression model.

sleep_model <- lm(
  depression_no_sleep ~ sleep_hours,
  data = analysis_data
)

summary(sleep_model)

Regression Results

All three models used the same 5,015 participants.

Sleep Measurement Coefficient P-value R²
Weekday -0.00857 0.7838 0.000015
Weekend -0.05464 0.0617 0.000696
Weekend–Weekday Difference -0.05365 0.0855 0.000590

Key findings:

  • All three coefficients were negative.
  • None of the models reached statistical significance.
  • All three R² values were extremely small.

The models explain less than 0.1% of the variation in depressive symptom scores.

Interpreting the Results

The weekday sleep regression model produced the following equation:

\[ \hat{Y} = 2.666 - 0.00857X \]

The coefficient indicates that one additional hour of weekday sleep was associated with an estimated decrease of approximately 0.009 points in depressive symptom score.

The 95% confidence interval for the weekday sleep coefficient was:

\[ -0.0713 \leq \beta_1 \leq 0.0502 \]

Because this interval includes zero, the results do not establish a statistically significant relationship.

Limitations

Several limitations should be considered when interpreting these findings:

  • Self-reported data: Sleep duration and depressive symptoms were reported by participants.
  • Modified depression score: One PHQ-9 question was excluded, creating a nonstandard eight-item score.
  • Observational data: The analysis cannot establish cause and effect.
  • Un-weighted sample: Survey weights and complex sampling design were not incorporated.
  • Limited model: Simple linear regression does not account for other factors such as age, stress, or lifestyle.

These limitations mean the findings should be interpreted cautiously.

Conclusion

This analysis examined the relationship between weekday and weekend sleep duration and reported depressive symptoms.

Main findings:

  • Weekday sleep showed almost no relationship with depressive symptom scores.
  • Weekend sleep and weekend sleep differences also showed very weak relationships.
  • None of the three regression models produced statistically significant results.

Final takeaway:

In this sample, sleep duration alone was not a meaningful predictor of reported depressive symptom scores.

Further research could examine additional factors that influence sleep and psychological health.

References