Transparencies_Topic_4

Transparencies_Topic_4 - A Random Effect in the Analysis...

Info iconThis preview shows pages 1–6. Sign up to view the full content.

View Full Document Right Arrow Icon
II Extra 1 4/6/2008 A Random Effect in the Analysis Need Not Imply A Random Draw in Data Generation The term "random effect" has come to be used more broadly than it was, say, 50 years ago. Analyses that are now described as including random effects apply to situations qualitatively different from those originally analyzed using random effects. As usage of “random effect” has broadened, fewer people seem to recognize a related distinction that has important consequences, both conceptual and practical. This lecture is about that distinction.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
II Extra 2 4/6/2008 Briefly, I'm going to make this argument: As analysis of richly-parameterized models become unified in the MLM framework (as in RWC), more analyses take the form of a random-effects analysis. An analysis including an effect with the form of a random effect does not necessarily imply that, out in the world, any random mechanism produced that effect. The form of the analysis should not be confused with the process in the world that produced the data . I make this argument using examples, most of which you’ve seen before.
Background image of page 2
II Extra 3 4/6/2008 Older, narrower notion of a random effect: -- A random effect represents draws from a population, either a real finite population or a hypothetical infinite one, and -- The draws are not of interest in themselves, but only as samples from the larger population.
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
II Extra 4 4/6/2008 Example 1: Draws from a real finite population . William Kennedy’s neurology lab uses new methods to count nerve fibers in skin and GI mucosa, an objective way to measure e.g., diabetic neuropathies. Recent dataset: 25 “normal” subjects; from each they took -- two types of skin samples: biopsies and blisters; -- each taken at two locations: calf and foot. The analysis involves three random effects: -- subjects, -- method-by-subject int (method = blister/biopsy); -- location-by-subject int (location = foot/calf). These random effects arise from sampling subjects. The analysis also includes a residual error = method- by-location-by-subject interaction.
Background image of page 4
4/6/2008 Example 2: Draws from a hypothetical infinite population . Dwight Anderson's lab in the dental school has done path-breaking studies working out, in molecule-by- molecule detail, the structure of the phi-29 virus. One of their measurement processes has three steps: -- producing, decomposing a batch of viral shells, -- separating the molecules by weight on a gel, -- burning blobs from each gel to give a measured weight for each blob. To get 98 measurements for one molecule, they used -- 9 batches, 11 gels, 7 oxidizer runs, -- 98 measurements grouped irregularly into batches, gels, and oxidizer runs. I treated batches, gels, and oxidizer runs as three
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 6
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 24

Transparencies_Topic_4 - A Random Effect in the Analysis...

This preview shows document pages 1 - 6. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online