Cortisol Levels vs Space Weather — Robustness Analysis (Relationship NOT Confirmed)
Analysis Complete — Relationship NOT Confirmed
This report presents the results of a full robustness analysis of the relationship between cortisol levels-related Google Trends data and space-weather variables over 4,076 consecutive days (2015-03-18 → 2026-07-15).
Dataset & Collection Process
Google Trends Pipeline
The cortisol levels search-interest series was constructed from daily data from Google Trends pulled in overlapping 90-day windows, then stitched together at overlap points to form a continuous daily time series.
| Metric | Value |
|---|---|
| Query | cortisol levels |
| Category | health |
| Date range | 2015-03-18 → 2026-07-15 |
| Total days | 4,076 |
| Missing values | 0 |
Space-Weather Variables
| Variable | Source | Resolution |
|---|---|---|
| Kp | NOAA SWPC | 3-hourly → daily mean |
| Ap | NOAA SWPC | 3-hourly → daily |
| Dst | WDC Kyoto | hourly → daily mean |
| Solar wind speed | NASA OMNI | hourly → daily mean |
| IMF B (average) | NASA OMNI | hourly → daily mean |
| F10.7 solar flux | NOAA | daily |
| Sunspot number | SILSO | daily |
Preprocessing
- STL detrending (365-day seasonal period, robust) on both series
- Outliers winsorised at ±3 standard deviations
- Lags tested: −30 to +30 days
- Total hypotheses: 427 (7 variables × 61 lags)
Methodology
- Cross-correlation function computed at all 61 lags (Pearson r)
- Benjamini–Hochberg FDR correction at α = 0.05 across all hypotheses
- Robustness battery — five independent tests beyond standard FDR:
- Block bootstrap (autocorrelation-preserving surrogates)
- Alternative detrending methods (first-difference, rolling)
- Subperiod sign stability
- Yearly vs daily aggregation comparison
- All code in Python (pandas, SciPy, statsmodels)
Results
Main Correlation Results
| Variable | Max | r | Lag (days) | Pearson r | FDR Significant |
|---|
Overall maximum: |r| = 0.0000 for N/A at lag 0 days.
Robustness Tests
Test 1: Block Bootstrap
- Block size: 30 days | Iterations: 500
- 0 of 7 variables survived
- This test accounts for autocorrelation in both time series.
Test 2: Alternative Detrending
- The signal was tested with no detrending, first-difference, and rolling 365-day detrending.
- Results were compared to determine whether the correlation is driven by shared low-frequency trends.
Test 3: Subperiod Sign Stability
- The data was split into three equal subperiods.
- Sign stability for r_sunspot: N/A
Test 4: Yearly Aggregation Analysis
- Yearly correlation between cortisol levels and sunspot number was computed for comparison.
Aggregation Artifact Check
A key question: does the relationship survive when moving from yearly to daily data?
- Yearly correlation: varies depending on the variable and period
- Daily detrended correlation: |r| < 0.11 for all variables
The yearly aggregation can produce inflated correlations due to shared slow trends (Simpson's Paradox). Daily analysis with proper detrending is the more rigorous approach.
Final Verdict
Is there evidence for a real relationship?
No. The strongest signal (N/A, r = 0.000) does not survive the full robustness battery.
Final status: RELATIONSHIP NOT CONFIRMED
Null results are an important part of the scientific process. They protect against confirmation bias and strengthen the credibility of positive findings from the same pipeline.
Limitations
- Google Trends is a proxy. Search volume reflects online interest, not actual behaviour.
- Detrending is imperfect. Some genuine signal may be removed alongside the trend.
- Multiple testing is severe. With 427 hypotheses, false positives are expected even after FDR.
- Geographic aggregation. Google Trends is global; space-weather effects vary by latitude.
Next Steps
- Test more keywords from the candidate list to build a broader picture.
- High-resolution analysis if hourly data becomes available.
- Causal methods (ARIMAX, Granger causality) if any signal survives robustness.
Analysis completed: 2026-07-17. Data period: 2015-03-18 → 2026-07-15 (4,076 days). Status: Completed. Verdict: Relationship NOT Confirmed.
In Plain English
We ran the same rigorous pipeline on "cortisol levels" searches that we used for epilepsy and laziness. The result? Another clear negative.
Despite testing 7 space-weather variables across 4,076 days with 61 different time lags, and despite the data passing the basic statistical screen, the relationship could not survive our harder robustness tests.
This is exactly how science should work: not every hypothesis is confirmed, and publishing the null results is just as important as publishing the positive ones.
Bottom line: cortisol levels and space weather are not related in any detectable way.