STA 2053 Projects

These will be due at the end of exam week. Exact details will be provided later. I'm expecting 5-10 typeset pages, maybe with an appendix showing computer input and output. I don't expect this to take more time than about two homework assignments. Be careful not to set yourself a task that is too ambitious. I am very available to consult. I hope I can help you keep it fairly small.

Here are some potential topics.

• Your idea. Please let me know what it is in advance if possible.
• Categorical with lavaan - explain and give a new example with simulated data. Do not even think about summarizing all of what one could do with different types of categorical data. One model for one kind of data is enough.
• Connection of SEM to neural networks, maybe deep learning -- in the language of statistics.
• Can SEM help with raw PREDICTON? Discuss and seek an example with simulated data.
• Dehan Kong, Shu Yang and Linbo Wang. (2021+). Identifiability of causal effects with multiple causes and a binary outcome. Biometrika, to appear. Get a copy of the paper and explain what they did.
• Judea Pearl and contemporary causal modelling. One paper is enough. It does not have to be by Pearl himself. It could be an attack on him, or a review of related papers.
• Random starting values (for lavaan).
• Analyze some real data. I have been told there's a lot out there, but I have not really investigated this. I do have one data set, based on a set of twins. The reference is

  Clark, P. J., Vandenberg, S. G., and Proctor, C. H. (1961), On the relationship of scores on certain psychological tests with a number of anthropometric characters and birth order in twins, Human Biology, 33, 163-180.

  The variables are

  • Sex
  • Identical or fraternal
  • Progressive Matrices Test
  • Chicago Test of Reasoning Ability
  • Chicago Test of Yerbal Ability
  • Total Test Score
  • Head Length
  • Head Breadth
  • Bizygomatic Breadth, basically how far apart the eyes are
  • Head Circumference
  • Height
  • Weight
  • Cephalic Index
  • Index of Cross- Sectional Head Area
  • Reciprocal Ponderal Index (like BMI)
  • Birth Order
  If you want to give this a try it's okay, but let's discuss the model. Link to the data set is HERE.
• I will add more possibilities if I think of some.