Linear regression residual net profit
factor.formula
in:
- :
represents the i-th quarter, where i is from the most recent quarter (t) back to N quarters ago, i.e. i = t, t-1, t-2 ... t-N+1
- :
Indicates the number of recent quarters used for regression analysis. The default value is 8 and can be adjusted according to actual conditions.
- :
Represents the net profit attributable to the parent company in the i-th quarter. This data needs to be normalized by Z-score to eliminate the differences in dimensions and distribution.
- :
Represents the non-operating income of the i-th quarter. This data needs to be normalized by Z-score to eliminate the dimension and distribution differences.
- :
Represents the cash paid to and for employees in the i-th quarter. This data needs to be normalized by Z-score to eliminate dimensional and distribution differences.
- :
The intercept term of the regression model indicates the expected value of the dependent variable when the independent variable is 0. It is not used directly in factor calculation and is only used for regression model construction.
- :
The coefficient of non-operating income in the regression model indicates the impact of each unit change in non-operating income on net profit when other factors remain unchanged.
- :
The coefficient of cash paid to and for employees in the regression model indicates the impact of each unit change in cash paid to and for employees on net profit, assuming other factors remain unchanged.
- :
Represents the residual term of the regression in the i-th quarter, reflecting the part of net profit that is not explained by non-operating income and cash paid to and for employees, that is, the purified net profit of the current period. The value of this factor is the residual corresponding to the most recent quarter (t), denoted as $\epsilon_0$
factor.explanation
Financial data contains both effective information that can predict future stock prices and noise that has no predictive power for stock prices. Improving the signal-to-noise ratio of data is the key to constructing effective factors. Net profit is affected by many factors, some of which may be weakly related to the company's core operating capabilities, such as non-operating income and cash paid to employees. This factor aims to eliminate these noises through linear regression, thereby improving the predictive power of net profit. Specifically, through the regression model, we try to find the part of net profit that can be explained by non-operating income and cash flow paid to employees, and treat it as noise elimination. The remaining residual is considered to be a signal that is more related to the company's core profitability. Therefore, this factor is named "Linear Regression Residual Net Profit". Through this method, a purer net profit signal can be obtained, thereby improving the effectiveness of factor stock selection. Z-score standardization processes all variables before regression in order to eliminate the differences in dimensions and distribution between different variables, making regression analysis more reasonable and reliable.