Correlations

Overview

Our correlation analysis system examines the relationship between users created Timeseries and stock market behavior. It aims to identify potential connections between central bank statements and various stock metrics, including price movements, returns, and volatility.

Key Features

Multi-timeframe Analysis: The system analyses correlations across different time periods, including same-day, next-day, next-week, and next-month intervals.

Comprehensive Stock Metrics: Beyond just price, the analysis considers stock returns and volatility, providing a more nuanced view of market behavior.

Statistical Rigor: The system employs statistical methods to ensure the reliability of the correlations found, including sample size considerations and confidence bounds.

Strength Classification: Correlations are classified as weak, medium, or strong, allowing for easy interpretation of results.

Confidence Scoring: Each correlation is assigned a confidence score, indicating the reliability of the finding.

Data Visualization: The results are formatted for easy visualization, including relevant stock data for correlated tickers.

Correlation Types

The correlation analysis calculates several types of correlations between user ratings and stock performance:

1. NextDayPrice: Correlation between the user rating and the stock's price on the next trading day.

2. SameDayReturn: Correlation between the user rating and the stock's return on the same day.

3. NextDayReturn: Correlation between the user rating and the stock's return on the following trading day.

4. NextWeekReturn: Correlation between the user rating and the stock's return over the next 5 trading days (approximately one week).

5. NextMonthReturn: Correlation between the user rating and the stock's return over the next 21 trading days (approximately one month).

Each correlation type measures how well the user ratings relate to or predict stock performance over different time horizons. A positive correlation suggests that higher ratings tend to be associated with higher prices or better returns, while a negative correlation suggests the opposite.

The strength of the correlation (weak, medium, or strong) and its statistical significance are also calculated to help interpret the results.

Correlation Strength

The strength of correlations is categorised as follows:

• Insignificant: Absolute correlation value less than 0.3 (only for sample sizes of 30 or more)

• Weak: Absolute correlation value between 0.3 and 0.4

• Medium: Absolute correlation value between 0.41 and 0.60

• Strong: Absolute correlation value 0.61 or higher

Note: For sample sizes between 10 and 29, correlations are categorised as weak (< 0.4), medium (0.4 - 0.60), or strong (> 0.60). Sample sizes below 10 are marked as "insufficient data".

These categories help interpret the practical significance of the correlations found between user ratings and stock performance metrics.

Process

After a timeseries has been created, a procsess is started to look for correlations with this data. This process can take a few minutes so be sure to check back on your timeseries.

1. The system retrieves the timeseries data and stock market data for a specified question or time period.

2. It computes correlations between the users timeseries and various stock metrics.

3. The correlations are analysed for strength and significance.

4. Results are summarized, highlighting the most significant findings.

5. The data is formatted for visualization and further analysis.

The specific algorithms and thresholds used in the correlation analysis are proprietary and continuously refined to improve accuracy and insights.