I came across the term "pretest estimator" in an econometrics book. Could someone elaborate on it?
I assume that what they are referring to is that the pre-intervention characteristics are used in the regression model as covariates (independent variables). Thus the outcome (Y) is regressed on these pre-intervention (pretest) covariates as a means of controlling for potential confounders.
This term could also be referring to the actual pretest value (baseline level of a test/outcome prior to the intervention), which would be used as an adjuster in a difference-of-differences model (post - pre)tx - (post - pre)control.
Please provide the name of the book and page that describes what you are referring to.
Ariel
If it is a baseline measure on the same scale as the outcome, there is much discussion about the (post-pre) versus ANCOVA approaches. Wainer (1991, Psych Bull) has a nice discussion using Rubin's Casual Model (RCM). The attached paper relates to how measurement error (and other things) can affect the choice.
Article Comparing groups in a before-after design: When t test and A...
Thank you very much for answering. Sorry for the vague question.
The book : Greene 6e , Econometric Analysis , Page 135.
This is not a full reference. What is Greene's full name, what year was this published, and by whom?
I found an older version of the book online and tracked down the term "pretest estimator"
Below is relevant paragraph. I don't agree with the logic in creating this third term, and apparently Greene doesn't either.
"Faced with a variable that a researcher suspects should be in their model, but which is causing a problem of collinearity, the analyst faces a choice of omitting the relevant variable or including it and estimating its (and all the other variables’) coefficient imprecisely. This presents a choice between two estimators, b1 and b1.2. In fact, what researchers usually do actually creates a third estimator. It is common to include the problem variable provisionally. If its t ratio is sufficiently large, it is retained; otherwise it is discarded. This third estimator is called a pretest estimator. What is known about pretest estimators is not encouraging. Certainly they are biased. How badly depends on the unknown parameters. Analytical results suggest that the pretest estimator is the least precise of the three when the researcher is most likely to use it. [See Judge et al. (1985).]"
Thank you! yes, this is exactly the paragraph. is my understanding correct? " By pretest estimator Prof Greene means the estimate of coefficient of the variable which is suspected to cause problem such as multicolinearity and is kept in the model when its t stat is significant"
No, it is a third variable that is generated that is a combination (interaction term) of the two problematic variables (the two variables that are colinear).
- Badges
- Science topic
- Similar topics
- Books