All entries for Friday 12 August 2005

August 12, 2005

IPW continues

About unweight-M-estimator;
(1) Should note first that the weights we are talking here are probabilitiy weights, not the weights that are provided by the survey conductors.

(2) the result of Assumptions 5.1 and 5.2 in Wooldridge (2002) coincides with that from the missing-data workshop in London. That is, the unweight M-estimator (complete case analysis) is valid ( unbiased and consistent) when R depends on X or is independent of both Y and X. (X here must be observable (not a latent variable like ability) but is allowed to be missing when R=0)

Missing-data workshop ( the introductory one) maintains that if R depends on X, the adjusted mean of Y for X is unbiased. ( adjusted mean is the coefficient of that X in a regression model) Also, if R depends on another random variable, say, S where S is independent of Y and X, then the result is also unbiased. Note that, in their example, S is uniformly distributed in (0,1). Can S be normally distributed???

Note that, from this course, the complete-case analysis is biased if R depends on Y or on both Y and X.

Wooldridge (2002) adds another condition to the conditions mentioned above. He shows that correct specification of the distribution of Y|X or that of mean function is also required.

August 2005

Mo Tu We Th Fr Sa Su
Jul |  Today  | Sep
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30 31            

Search this blog

Favourite blogs


Most recent comments

  • Hi im yo man my intenet namee imm 43 i woork in the toilet withh laptops this coool webbsete by yo man on this entry
  • this suck i'm out of here by katie on this entry
  • this is a stupid website im 89 years old in two weeks it's my birthday how old would i be this is an… by glf on this entry
  • What r u talking about? I think u must be confused. No No No ur major is not econometrics. U have b… by MMP on this entry
  • This blog is temporary out of service because the owner is still busy watching TV (no matter what la… by MMP on this entry

Blog archive

Not signed in
Sign in

Powered by BlogBuilder