DataModel#

class qf_lib.common.utils.factorization.data_models.data_model.DataModel(data_model_input: DataModelInput)[source]#

Bases: object

Class grouping the results of factorization.

Parameters:: data_model_input – data from which the model is built

AUTOCORR_MAX_LAG = 3#: int maximal lag used during testing for autocorrelation of the fit; lags used for testing will be values 1, …, autocorr_max_lag

AUTOCORR_SIGNIFICANCE_LEVEL = 0.05#: float significance level for the autocorrelation of the fit test

autocorrelation#: Extension of Durbin-Watson test to add many lags (1-5). 0 - not autocorrelated, 1 - autocorrelated.

coefficients#: Vector of coefficients [beta1, beta2, …].

condition_number#: Condition number of a matrix measures the sensitivity of the solution of a system of linear equations to errors in the data.

cooks_distance_tms#: Cooks distance. Used for checking the influence of outliers for the model.

durbin_watson_test#: Used to test if linear regression residuals are uncorrelated. Small p-values indicate correlation among residuals.

factors_performance_attribution_ret#: Vector containing annualised performance attribution of each factor.

fit_model#: Structure with a result of multilinear regression (based on all data points and using OLS to calculate coefficients).

fit_tms_analysis#: TimeseriesAnalysis class based on returns of the fit.

fitted_tms#: Fitted (predicted) response values based on input data.

fund_tms_analysis#: TimeseriesAnalysis class based on returns of the analysed fund.

heteroskedasticity#: Probability of a hypothesis that the error variance doesn’t depend on input data (regressors).

in_sample_and_out_sample_returns#: Returns of a fit based on in-sample coefficients. Vector with in-sample and out-of-sample simple returns. Its length is equal to length of fitted returns.

intercept#: Constant alpha (y = beta * x + constant).

ols_influence#: Class for calculating outliers and influence measures for OLS result.

oos_start_date#: Date on which the Out-Of-Sample period started (In-Sample vs Out-Of-Sample test).

r_squared_of_each_predictor#: Concerns about collinearity can be ignored if rSquare is higher than rSquare of each predictor.

risk_contribution#: Vector containing normalised risk contribution of each factor.

unexplained_performance_attribution_ret#: Scalar with annualised return unexplained by factors.