Schedule: click on 'schema' in the column to the left
Exercises
Excercise 1.
Excercise 1. Gunnar Gelin.
Excercise 2.
singers -Excel file -->>
samplesingers -Excel file -->>
The deadline has passed
Excercise 3.
<
Excercise 3. Answer.
Excercise 4.
The deadline has passed.
Course work
Course work
Data 1.
Data 2.
Lecture notes
Module 1, Science
Module 2, Models for Measurement Errors
Module 2, Models for Measurement Errors continued
Module 2, LCA
Module 3: Observational studies 1
Module 3: Observational studies 2
Other texts to read
Biemer's example on reliability of marijuana use questions
Brief overview
The course starts by discussing science and the relationship between science and statistics. We will then move over to issues in statistics including:
- Are there 'objective' methods at all in statistics?
- What are the benefits of using statistical models, even if the model is not true?
- What are the benefits with asymptotic reasoning, although in reality no sample sizes approach infinity?
- In poststratification, should the poststratum sample sizes be viewed as random or fixed (they are random, but should they be viewed as random)?
- In science, very often the most interesting variables are those that cannot be observed. Is it scientific to talk about things that you cannot even see?
Particular issues that we will discuss include:
- How to treat some random sample sizes, such as the parts of the sample that happen to fall in poststrata
- Measurement error models in surveys (it is very rarely possible to observe a measurement error directly)
- Some uses of latent variables
- Use of models in surveys
- One particular example of model misspecification that gave rise to an animated debate
- Are frequentist methods objective (if not, why quibble about Bayesian inference?)
- What is so sacred about randomness, do we need random sampling?