FAQ
Corrlens Correlation Stats: What They Mean
This page explains the statistics shown in Corrlens, how they are calculated, and how to interpret them.
What stats are available?
Correlation (r): Pearson correlation coefficient between aligned values.
Overlap Points: Number of paired observations used in the calculation.
Alignment Mode: Daily, weekly, or monthly bucketing used before pairing.
Date Range Used: First and last date from the aligned overlap window.
Dropped Counts: Invalid, duplicate, and alignment-dropped points for each series.
How is correlation calculated?
r = cov(X, Y) / (std(X) * std(Y))
- Load two time series (FRED or stock).
- Normalize timestamps and clean invalid values.
- Deduplicate by date (last value wins).
- Align both series by selected mode (daily/weekly/monthly).
- Compute Pearson correlation on aligned pairs only.
How to interpret r
Positive (r > 0)
Series tend to move in the same direction.
Negative (r < 0)
Series tend to move in opposite directions.
Near 0
Little linear relationship in the selected window.
Near ±1
Strong linear relationship (same/opposite direction).
Why correlations are useful
Spot shared trends faster: Correlation helps you quickly identify assets and macro series that tend to move together.
Find similarities across markets: You can compare stocks, FRED indicators, or mixed sets to reveal patterns that are not obvious from one chart alone.
Improve research focus: High or changing correlations can highlight which relationships are worth deeper investigation.
Monitor regime changes: When previously related series decouple, that can signal shifts in market behavior or macro conditions.
Important caveats
- Correlation is not causation.
- Results depend heavily on selected window, alignment mode, and data availability.
- Structural breaks and regime changes can make historical relationships unstable.
- Very low overlap counts can produce noisy and misleading correlations.