AlgorithmConsider the linear regression with samples satisfying:
Goal: We aim at ex post statistical inference on individual model parameters , in terms of frequentist p-value and confidence interval. MethodBelow we provide a brief explanation of the online debiasing method. For more details and scissions, we refer to our paper. Denote by the Lasso estimator
The online debased estimator takes the form
How to choose decorrelating matrices?We focus on times series as an important application of adaptively collected data and consider a vector autoregression (VAR) model for time series. By a proper change of variables, a Var(d) model can be represented as the linear regression with parameters and sample size where is the time horizon where we observe the times series, and is the lag. Our primary interest is in high-dimensional Var(d) model, where the number of model parameters exceeds the sample size . The proposed online debiasing provides valid statistical significance measures for the model parameters (the time invariant matrices in the Var(d) model). To construct decorrelating matrices that satisfy the predictability condition, we proceed as follows:
1. For do
Construct by solving the following optimization:
With the standard basis element with one at the -th position and zero everywhere else. 2. Set . In words, stack the constructed vectors as rows of .
Our analysis in the paper suggests the choice of episode lengths . The code can take the parameter as an input. |