Inferring serial correlation with dynamic backgrounds
Authors: Song Wei, Yao Xie, Dobromir Rahnev
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical simulation and a real study in psychology demonstrate the excellent performance of our proposed method compared with the state-of-the-art. Extensive numerical experiments are performed to validate the effectiveness of our proposed method. We also test our method on a real psychology dataset |
| Researcher Affiliation | Academia | 1School of Industrial and Systems Engineering, Georgia Tech, 2School of Psychology, Georgia Tech, Atlanta, GA, USA. Correspondence to: Yao Xie <yao.xie@isye.gatech.edu>, Song Wei <song.wei@gatech.edu>. |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any concrete access to source code for the methodology described, nor does it state that code is released or available in supplementary materials. |
| Open Datasets | Yes | The data are taken from a publicly available database introduced by Rahnev et al. (2020) with 149 individual datasets with human data on different tasks. |
| Dataset Splits | Yes | Remark 2 (Cross-validation for parameter tuning). We now performed a type of leave-one-out cross-validation (CV). Note that we cannot use vanilla CV because of the presence of the dynamic, unstructured background and serial correlation, as explained above. Instead, in each trial, we leave out one observation and treat it as missing-value (we have to deal with it anyways using imputation in real-data). We then fit the model and obtain one CV error using the predicted value for the left-out observation. We repeat these T times, where T is the length of the sequence, to obtain an estimate of the CV error. |
| Hardware Specification | No | The paper does not provide any specific hardware details (such as GPU/CPU models, memory, or cloud instances) used for running its experiments. |
| Software Dependencies | No | The paper states 'The convex program (2) is solved by the cvx package (Grant & Boyd, 2014)', but it does not specify a version number for the 'cvx package' or any other software dependencies. |
| Experiment Setup | Yes | We adopt the following experimental setting: α1 = 0.1, σ2 0 = 0.1, T = 1000. The dynamic drift is piecewise constant with δ0 = 0.1, s = 100. The bootstrap replication is N = 100; we use standard normal random numbers as vt s in residual-based wild bootstrap (WB); for local block bootstrap (LBB), we choose block size b = 20 and local neighborhood length B = 50. |