LoadingLoading
Multicollinearity refers to the presence of a high degree of correlation between two or more independent variables in a multiple regression model.
Multicollinearity refers to the presence of a high degree of correlation between two or more independent variables in a multiple regression model. Multicollinearity can cause distorted or misleading results when a researcher or analyst attempts to determine the effect of each independent variable in a statistical model in predicting or understanding the dependent variable.
In general, multicollinearity can lead to wider confidence intervals, resulting in less reliable probabilities of independent variable effects.
In technical analysis, multicollinearity can lead to incorrect assumptions about investments. Typically, this occurs when multiple indicators of the same type are used to analyze a stock.
Statistical analysts use multiple regression models to predict the value of a specific dependent variable based on the values of two or more independent variables. The dependent variable is sometimes called the outcome, target, or criterion variable.
For example, a multiple regression model attempts to predict stock returns based on price-to-earnings ratios (P/E ratios), market capitalization, or other data. Stock returns are the dependent variable (outcome), while various financial data are the independent variables.
Multicollinearity in a multiple regression model indicates that collinear independent variables are not truly independent. For example, past performance may be correlated with market capitalization. Shares of companies that perform well are often accompanied by rising investor confidence, leading to increased demand for the company's shares, thereby increasing its market value.
Although multicollinearity does not affect regression estimates, it makes these estimates vague, imprecise, and unreliable. Therefore, it is difficult to determine the individual effects of the independent variables on the dependent variable. This will inflate the standard errors of some or all regression coefficients.
A statistical technique called the variance inflation factor (VIF) can detect and measure collinearity in multiple regression models. VIF measures the degree to which the variance of the estimated regression coefficient is inflated compared to the variance if the independent variables are not linearly related. A VIF of 1 means there is no correlation between the variables; a VIF between 1 and 5 means there is a moderate correlation between the variables; and a VIF between 5 and 10 means there is a high degree of correlation between the variables.
When analyzing stocks, you can detect multicollinearity by watching whether indicators plot the same graph. For example, selecting two momentum indicators on a trading chart will often generate trend lines showing the same momentum.
Multicollinearity may occur when two independent variables are highly correlated. Multicollinearity can also result when an independent variable is calculated from other variables in the data set, or when two independent variables provide similar and repeated results.
If you create two or three trading indicators of the same type using the same data, the results will be multicollinear because the data and the way it is manipulated are very similar.
IMPORTANT: Statistical inferences generated by models containing multicollinearity may be unreliable.
Perfect collinearity means that there is an exact linear relationship between multiple independent variables, and usually you can see in the chart that the data points fall completely on the regression line. In technical analysis, this can also occur when two indicators that measure the same indicator (such as volume) are used simultaneously. If you superimpose one indicator on another, there will be no difference.
High collinearity means that there is a certain correlation between multiple independent variables, but this correlation is not as close as complete collinearity. Not all data points fall on the regression line, but it still indicates that the correlation between the data is too close to be used.
In technical analysis, indicators with high collinearity often produce very similar results.
Structural collinearity occurs when you use data to create new features. For example, if you collect data and use it in other calculations and then perform a regression analysis on the results, the results will be related because they are derived from each other.
This is a common type of collinearity in investment analysis because the same data is used to create different indicators.
Poor experimental design or data collection procedures, such as using observational data, often results in data-based collinearity, in which some or all variables are correlated.
The stock data used to create indicators is typically collected from historical prices and trading volumes, so there is less chance of multicollinearity due to poor data collection methods.
Multicollinearity is a common consideration when investing, especially when performing technical analysis to predict future price movements for a security, such as a stock or commodity futures.
Market analysts want to avoid using collinear technical indicators that are based on very similar or related inputs; these inputs do not refer to the data itself, but how the data is manipulated to achieve results.
Therefore, the analysis must be based on significantly different indicators to ensure that the market is analyzed from an independent analytical perspective. For example, momentum indicators and trend indicators share the same data, but they will not be perfectly collinear and may not even be highly collinear. These two indicators produce different results depending on how the data is manipulated.
Note: Most investors will not pay too much attention to the data and technology behind the calculation of indicators - it is enough to understand what multicollinearity is and how it affects the analysis.
One common approach to eliminating multicollinearity problems is to first identify the collinear independent variables and then remove one or more of the independent variables. Typically, in statistics, calculations of variance inflation factors are performed to determine the degree of multicollinearity. Another way to address multicollinearity is to collect more data under different conditions.
Renowned technical analyst John Bollinger, the creator of the Bollinger Band indicator, once wrote: "A basic rule of successful use of technical analysis requires avoiding multicollinearity among indicators." To solve this problem, analysts avoid using two or more technical indicators of the same type. Instead, they analyze a security using one type of indicator (such as a momentum indicator) and then conduct a separate analysis using another type of indicator (such as a trend indicator).
For example, Stochastic, RSI, and Wm%R are all momentum indicators that rely on similar inputs and may produce similar results. In the above image, the Stochastic and William indicators are identical, so using them together does not reveal much. In this case, it is better to eliminate one of the indicators and use one that does not track momentum. In the image below, Stochastic shows price momentum, while Bollinger Bandwidth shows price consolidation ahead of a price move.
To reduce multicollinearity in a statistical model, specific variables identified as most collinear can be removed. You can also try combining or transforming problem variables to make them less relevant. If this is still not feasible, there are improved regression models that better handle multicollinearity, such as ridge regression, principal component regression, or partial least squares regression. In stock analysis, the best approach is to choose different types of indicators.
Multicollinearity describes the relationship between variables, causing them to be correlated with each other. Multicollinear data pose problems for analysis because they are not independent.
When the variable inflation factor exceeds five, the data will exhibit a high degree of multicollinearity. If VIF is between 1 and 5, the variables are moderately correlated, while when it is equal to 1, it means there is no correlation. In technical analysis, the indicators are usually the same.
Perfect collinearity exists when there is an exact 1:1 correspondence between two independent variables in the model. The correlation coefficient can be +1.0 or -1.0.
Multicollinearity is a problem because it produces less reliable regression model results. This is because a wider confidence interval (increased standard error) may reduce the statistical significance of the regression coefficient. In stock analysis, it can lead to erroneous impressions or assumptions about investments.
Multicollinearity exists in multiple regression equations where the independent variable is highly correlated with multiple other independent variables. Multicollinearity is a problem because it makes statistical inference less reliable. However, the variance inflation factor (VIF) can provide information about which variables are redundant, so variables with high VIF can be removed.
Multicollinearity becomes a problem when using technical analysis because many indicators present data in the same way. To avoid this, it's better to use indicators that don't measure the same trend.
TheQuantHackers is a cutting-edge community composed of a group of people from the world's top universities and quantitative finance practitioners, dedicated to exploring the infinite possibilities in the fields of artificial intelligence (AI) and quantification (Quant). Our team members come from world-renowned universities such as the University of Cambridge, Oxford University, Harvard University, ETH Zurich, Peking University, and the University of Science and Technology of China. Our external consultants come from first-class companies such as Microsoft, HSBC, Citadel, Man Group, Citi, Jump Trading, and top domestic private equity companies.
Upgrade to unlock all institutional-grade algorithms, derivative pricing engines, factor backtesting frameworks, and live QuantLab execution.
Explore Membership Access
Pre-order · £490Quantitative analytics platform
Waitlist closed — signal locked. Pre-order TQH TERMINAL now at £490 with launch-day delivery.