STA248H1: Statistics for Computer Scientists

Course Description

A survey of statistical methodology with emphasis on data analysis and applications. The topics covered include descriptive statistics, data collection and the design of experiments, univariate and multivariate design, tests of significance and confidence intervals, concept of power, multiple regression. Students learn to use a statistical computer package as part of the course (Note: STA248H1 does not count as a distribution requirement course).

Programming languages

Both R and Python are acceptable computing languages in this course, we will see use of both throughout the course. Make sure you master at least one of them by end of this course. You will complete at least two of the specified DataCamp Programming modules as part of this course.

Course Website

Everything you need for the course is hosted here. Please check frequently for posting of lecture slides, practice problems and test information.

Textbook and learning resources

There is no official textbook for this course, but if you wish to have additional books/materials to look at, I suggest the following (all available through U of T library either as a physical copy or electronically).

Open Intro Statistics, 3rd edition available for download here
Chapter 2-3 can be used as review for STA247
topics in Chapter 1,4,5,7 and some of 8 are covered in this course
Probability and Statistics for Engineering and the Sciences 9th Edition by Jay L. Devore

Getting help

Please see website for schedule of office hours. Also, it is not a bad idea to google - and learning to use it effectively will become one of your most valuable skills in your future study/work.

Note that we will have additional office hours before midterm and final exam. If you wish to meet/discussion outside of office hours please schedule with me at least one day in advance.

IMPORTANT: please include in the subject head the course code ``STA248’’ whenever you are emailing regarding this course.

Evaluation

Completion of two R/Python programming modules 10% (5% each)
Midterm 40%
Final Exam 50%

Course content

Lecture 1: Welcome and introduction to statistics (July 4th, Wednesday)

First half: course overview and connection with STA247 and STA302
Second half: summarizing data (graphically and through summary statistics)
Connection between statistics and parameters (ideas of a statistical model); population and sample.
descriptive statistics (idea of a centre/location, spread/scale of a distribution)
commonly used graphs (bar, histogram, boxplot, scatterplot, ecdf, qqplot)
graphical representation and information from the graphs (outliers, model assumptions, dependence)
data transformation

Lecture 2: Estimation - to think with data (July 9th, Monday)

First half: properties of estimates derived from data
unbiased, consistency
sampling distribution, standard error
method of moments
maximum likelihood estimates
Second half: idea of inferential learning; bootstrap sampling
confidence intervals (CI)
parametric and non-parametric bootstrap
CI for \(\mu\) and \(\sigma^2\) (normal, t-distribution, or binomial distributions), what about the median?
bootstrap CIs.

Lecture 3: Hypothesis Tests and statistical significance

tests of significance
one-sided vs. two sided test
confidence intervals
power

Lecture 4: Testing: Univariate and multivariate design - I (July 11th, Wednesday)

formalize ideas of testing and inference
student’s t-test (one sample, two sample, and paired)
test of mean equality
assumptions
transformation

Lecture 5: Testing: Univariate and multivariate design -II (July 16th, Monday)

analysis of variance (one-way)
F-test in two samples
\(K\)-samples
know how to construct an one-way ANOVA table
Levene’s test (general idea for test of equality)
generalize the idea of F-test as a test for goodness of fit.
interactions (two-way - not tested)

Lecture 6: Summarize ideas of testing and inference (July 18th, Wednesday)

using bootstrap to estimate sampling distributions
multiple hypothesis (FDR and Bonferroni correction)
use and interpretation of p-value
quick review of ideas for midterm

Lecture 7: Midterm test (July 23th, Monday)

Location to be announced the week before.

Lecture 8: Simple and multiple regression (July 25th, Wednesday)

evidence for linear relationship
correlation coefficients
scatterplot and boxplot (when predictor variable is continuous or discrete)
a linear model
a simple linear model
model assumptions
idea of least squared method
measure of fit

Lecture 9: Midterm review (July 30th, Monday)

Tianle subbing in

Lecture 10: More on linear regression (August 1st, Wednesday)