This chapter begins with a discussion of the terminology and conceptual frameworks that are useful for contextualizing pre-modern Chinese sources about sex and sexuality. It then surveys several well-studied institutions and practices, including sex manuals, concubinage, female chastity, illicit sex, and literary representations of homoeroticism. The second half of the chapter reflects on three phenomena in works on the history of sexuality in pre-modern China, namely retrospective sexology, the censorship hypothesis, and the assumption of sex as a given. The author argues that while historians now no longer characterize sex culture in ancient China as either ‘liberated’ or ‘repressed’, as old sexologists did, we still tend to assume that the history of sexuality should primarily be about sexual practice and behaviour, despite the acknowledged lack of sources. The lack of sources, in turn, is often assumed to be the result of political and ideological censorship. More attention is needed to questioning scholars’ definition of the very subject matter, sex. The chapter concludes with a short review of scholarly approaches to comparing China with other cultures and a proposal of the ways in which a comparative history of sexuality can be productive.

Issues of joint dependence between the dependent variables and explanatory variables are discussed. Error components 2SLS or 3SLS estimator for an equation or a system is introduced. Identification conditions with prior restrictions on the parameters or the disturbance terms are illustrated with a triangular system.

Issues of some widely used nonlinear models such as the duration, count data, or nonparametric estimation of general nonlinear models in the presence of individual or time-specific effects are discussed. Consistent and asymptotically normally distributed estimators are introduced.

The advantages and challenges of using panel data are discussed. An outline of this monograph is given.

Dynamic linear error-components models are introduced. Issues of multi-dimensional asymptotics are discussed. Large sample properties of the instrumental variable estimator (IV), generalized method of moments estimator (GMM), and maximum likelihood estimator (MLE) when N is fixed and T goes to infinity, or T is fixed and N goes to infinity, or both N and T large, are examined. Bias correction estimators are introduced.

Advantages and challenges of using factor structure to control the impact of unobserved heterogeneity that varies across individuals and over time are discussed. Methods for the determination of the dimension of factor structure are also discussed.

Weak and strong cross-sectional dependence are discussed. Various spatial approaches to model cross-sectional dependence, such as the spatial error or spatial regressive approach for static or dynamic models, are introduced. Issues of endogeneity of spatial weight matrix, higher-order spatial weight matrix, and the mixed spatial and factor process are considered. The Lagrangian multiplier and CD tests for cross-sectional uncorrelation for various types of models are also discussed.

Issues of nonrandom sampling due to truncation, censoring, or sample selection rules are discussed in the presence of individual-specific effects. Symmetric trimming of sample observations to get rid of incidental parameters are introduced.

The use of a panel vector autoregressive model as a reduced form approximation to a panel dynamic system is introduced. Nonstationarity and cointegrations over time and across cross-sections are discussed. Identification conditions and MLE or GMM estimation of dynamic simultaneous equation models are considered.

Issues of quantile regression, simulation methods, multi-level panel data, errors of measurement, distributed lag models when T is short, rotating or randomly missing data, repeated cross-sectional data, and discretizing unobserved heterogeneity are discussed.

The motivation for generalizing unobserved heterogeneity of varying parameter models is discussed. Various fixed or random varying parameters across cross-sectional units and over time models together with their respective inference procedures are introduced from both the sampling approach and the Bayesian approach. Issues of correlations between parameter variation and regressors are also discussed.

Essentials of machine learning algorithms are briefly discussed. High-dimensional nonparametric inference and inference for low-dimensional parameters in the presence of big data are introduced. Decision-based prediction vs. causal based predictions and aggregate vs. disaggregate data analysis as well as issues of combining data from different sources are also discussed.