This paper extends three Lasso inferential methods, Debiased Lasso, $C(\alpha)$ and Selective Inference to a survey environment. We establish the asymptotic validity of the inference procedures in generalized linear models with survey weights and/or heteroskedasticity. Moreover, we generalize the methods to inference on nonlinear parameter functions e.g. the average marginal effect in survey logit models. We illustrate the effectiveness of the approach in simulated data and Canadian Internet Use Survey 2020 data.
翻译:本文将三种Lasso推理方法,去偏Lasso、$C(\alpha)$和选择性推理扩展到调查环境。我们建立了带有调查权重和/或异方差的广义线性模型中推断程序的渐近有效性。此外,我们将方法推广到非线性参数函数的推断,例如调查logistic模型中的平均边际效应。我们在模拟数据和加拿大互联网利用调查2020数据中展示了方法的有效性。