THE RESULT
What the data shows
The largest computed metric is 0.927 for co_gt / pt08_s2_nmhc; the smallest is -0.8112 for co_gt / pt08_s3_nox. Metric: spearman_rho (Spearman correlation).
Paired sample sizes differ across references. Common environmental conditions can induce these associations.
| feature | spearman rho | absolute rho | paired rows | reference | pair |
|---|---|---|---|---|---|
| pt08_s2_nmhc | 0.927 | 0.927 | 7,344 | co_gt | co_gt / pt08_s2_nmhc |
| pt08_s1_co | 0.8794 | 0.8794 | 7,344 | co_gt | co_gt / pt08_s1_co |
| pt08_s5_o3 | 0.8557 | 0.8557 | 7,344 | co_gt | co_gt / pt08_s5_o3 |
| pt08_s3_nox | -0.8112 | 0.8112 | 7,344 | co_gt | co_gt / pt08_s3_nox |
| pt08_s4_no2 | 0.5944 | 0.5944 | 7,344 | co_gt | co_gt / pt08_s4_no2 |
| pt08_s5_o3 | 0.793 | 0.793 | 7,396 | nox_gt | nox_gt / pt08_s5_o3 |
| pt08_s3_nox | -0.7904 | 0.7904 | 7,396 | nox_gt | nox_gt / pt08_s3_nox |
| pt08_s1_co | 0.7248 | 0.7248 | 7,396 | nox_gt | nox_gt / pt08_s1_co |
| pt08_s2_nmhc | 0.7047 | 0.7047 | 7,396 | nox_gt | nox_gt / pt08_s2_nmhc |
| pt08_s4_no2 | 0.1602 | 0.1602 | 7,396 | nox_gt | nox_gt / pt08_s4_no2 |
| pt08_s5_o3 | 0.7086 | 0.7086 | 7,393 | no2_gt | no2_gt / pt08_s5_o3 |
| pt08_s3_nox | -0.6892 | 0.6892 | 7,393 | no2_gt | no2_gt / pt08_s3_nox |
THE METHOD
From source to answer
Pairwise-complete rank correlations across five sensor channels and three references.
The Python source exposes this study’s transformations. The complete project download includes shared preparation and evaluation routines.
THE NEXT DECISION
What follows from the finding
Investigate cross-sensitivity before assigning a sensor a single-gas interpretation.
Where the conclusion stops
The -200 sentinel is treated as missing. Reference instruments have incomplete coverage; comparisons use paired observations. Sensor calibration is retrospective, not a health or regulatory compliance assessment.
Related studies may reuse observations or holdouts. These are historical analyses; associations and backtests do not demonstrate commercial impact. Further model tuning needs new, untouched evaluation data.