The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

This labs' goal is for students to learn R commands and experience data with R.

This labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

The labs' goal is for students to learn R commands and experience data with R.

This lecture will review basic principles of Bayesian inference, Bayesian hierarchical models, and the BUGS tool for conducting Bayesian analyses.

Graphical Markov models use graphs with nodes corresponding to random variables, and edges that encode conditional independence relationships between those variables. Directed graphical models (aka Bayesian networks) in particular have received considerable attention. This lecture will review basic concepts in graphical model theory such as Markov properties, equivalence, and connections with causal inference.

This lecture will describe two applications of Bayesian graphic models.

The counterfactual approach to casual inference dates back at least to Neyman and has developed considerably in recent decades due to pioneering work by Rubin, Pearl, Robbins, and others. This lecture will introduce the counterfactual approach and discuss specific examples.

Estimating the effects of medical interventions from massive longitudinal observational medical data is of central importance in healthcare. This talk will describe typical methods used for this purpose and recent attempts to understand the limitation of these approaches.

This lecture will cover core Monte Carlo ideas such as Monte Carlo sampling, importance sampling, Markov chain Monte Carlo, and sequential Monte Carlo.

This lecture will introduce Bayesian model averaging and describe algorithms for applying model averaging in different contexts.

Algorithms for online (one example at a time) learning and model averaging.

This lecture will introduce various problems in the analysis of textual data such as text categorization, entity recognition, and authorship attribution, and describe related methodology.

Approximate Bayesian computation is an extraordinarily simple method for doing Bayesian calculations. The lecture will describe the basic approach and consider a number of applications.

In this lecture, we will discuss basic experimental design principles in data collection and issues regarding data quality. Specific data examples such as the entron data set will be used.

This lecture will cover data summarization and visualization tools such as kernel estimation, loess, scatterplot and dimension reduction via principal component analysis (PCA). Specific data examples will be used.

This lecture reviews least squares (LS) method for linear fitting and its statistical properites under various linear regression model assumptions. Methods will be illustrated with real data examples from instructor's research projects.

This lecture will generalize LS to weighted LS (WLS) and use WLS to connect with generalized linear models including logistic regression. Remote sensing data for cloud detection will be used.

LS and Maximum Likelihood estimation (MLE) overfit when the dimension of the model is not small relative to the sample size. This happens almost always in high-dimensions. Regularziation often works by adding a penalty to the fitting criterion as in classical model selection methods such as AIC or BIC and L2-penalized LS called Tikhonov regularization or Ridge regression. We will also introduce Cross-validation (CV) for regularization parameter selection.

Boosting and Support Vector Machines (SVMs) are two successful supervised machine learning methods. This lecture will introduce them and relate them to LS and generalized linear models. These methods will be applied to the remote sensing data problem.

This lectures will cover two modern regularization topics: (1) Least Absolute Shrinkage and Selection Operator (Lasso) as a convex relaxation to AIC or BIC and (2) low-rank regularization arising from the Netflix competition. A subset of the netflix data will be investigated.

This lecture will discuss variants and extenstions of Lasso such as Lasso+LS, adaptive Lasso, and group Lasso.

This lecture will give a heuristic overview of theoretical results on Lasso that explains when and why Lasso and extensions work.

This lecture will illustrate the power of the sparse coding principle and low-rank regularization in modeling neuron responses to natural images in the very challenging visual cortex area V4.