臺灣博碩士論文加值系統

科學研究顯示，睡眠習慣和一個人的健康有很大關係，但是什麼
樣的睡眠習慣比較好，一直都沒有定論。本論文藉由知名作家的睡眠
資料找出在這組特定樣本下起床時間和生產力的關係。最小平方估
計法(OLS) 顯示起床時間和生產力有負向關係。但當考慮到內生性問
題，兩階段最小平方法(2SLS) 的結果則顯示未有任何相關。在分量迴
歸之下估計法之下，起床時間對作家生產力的條件分配有著異質性影
響，在低分量的分配部分，負相關較有統計上顯著。然而在實際上，
真正重要的變數和模型的函數形式通常是未知的，我們使用double
selection 的步驟解決上述問題，並控制內生性偏誤。

Science often says that sleep plays an important role in good health and
well-being throughout one’s life. However, what kind of sleep patterns is
good to human beings? In this paper, given the data of sleep habits of some
historical figures — famous writers, we would like to find the relation between
wake-up time and productivity. The OLS estimates reveal that there
is a negative relationship between them. Taking endogeneity problem into
consideration, the 2SLS method is applied, which does not report any relation
between wake-up time and productivity. According to the regression
quantile estimates, wake-up times have heterogeneous effect over the conditional
distribution of writers’ productivity. In practice, since the relevant
variables and the correct functional form are usually unknown, the method
of LASSO can be modified, namely the double-selection method, cf. Belloni,
Chernozhukov and Hansen (2014), to cope with variable selection and
model specification under endogeneity. We demonstrate the double-selection
method in the present study. Quantile treatment effects are investigated as
well.

口試委員會審定書i
誌謝ii
摘要iv
Abstract v
1 Introduction 1
2 Data Description 3
3 Econometric Methods under Exogeneity 8
3.1 Ordinary Least Squares . . . . . 12
3.2 Post-LASSO . . . . . 17
3.3 Conditional Quantile Regression . . . . . 22
4 Econometric Methods for Endogeneity 24
4.1 Double-Selection with Post-LASSO . . . . . 24
4.2 Triple-Selection with Post-LASSO . . . . . 28
5 Discussion and Conclusion 44
References 46

[1] brainpickings, Famous Writers'' Sleep Habits vs. Literary Productivity, Visualized.
https://www.brainpickings.org/2013/12/16/writers-wakeup-times-literaryproductivity-visualization/

[2] Gareth James, Daniela Witten, Trevor Hastie and Robert Tibshirani. An Introduction to Statistical Learning: With Applications in R. New York: Springer, 2013.

[3] Alexandre Belloni, Victor Chernozhukov and Christian Hansen. High-dimensional methods and inference on structural and treatment effects. Journal of Economic Perspectives, 28(2):29–50, 2014.

[4] Joshua D. Angrist and Jörn Steffen Pischke. Mostly Harmless Econometrics: An Empiricist''s Companion. Princeton University Press, 2008.

[5] Jeffrey M. Wooldridge. Econometric Analysis of Cross Section and Panel Data. MIT Press, 2002.

[6] Hlavac, Marek. stargazer: Well-Formatted Regression and Summary Statistics Tables. R package version 5.2. http://CRAN.R-project.org/package=stargazer, 2015.