This assumption may not hold, so that the MI estimate of mean CVL in any given year remains biased.
Thus, the usual MI estimate of θ is expressed by: θ ^ MAR = 1 M ∑ m = 1 M θ ^ m. (see Appendix for its estimated variance).
Finally, of the 103 papers reviewed, only 3 papers (3%) performed a sensitivity analysis following MI to investigate the robustness of the MI estimate to departures from the MAR assumption.
It is accordingly difficult to evaluate the influence of a mis-estimate of 0.1 ~ 0.2 in magnitude.
The recently revealed mis-estimate of the war budget amounts to ten billions of dollars for a single year.
MI estimates of NRI and IDI indexes were close to those obtained with the full cohort analysis and did not depend on the subcohort size.
By accounting for the variability between the completed (imputed) datasets, MI produces a valid estimate of the precision of the final MI estimate.
The estimates were performed with the help of Stata's class of - mi estimate - commands.
Logistic regression models were used in the case control sample, comprising MI cases and population controls, to assess the relative risk of developing MI, estimated as ORs.
The estimated variance of the overall MI estimate allows for within-imputation (i.e. the uncertainty in the estimate within each completed dataset) and between- imputation (i.e. the uncertainty between the estimates across the completed datasets) variability [ 15, 16].
The MI estimate, also unbiased, is the mean of the M estimates: (1) θ ^ M I = 1 M ∑ m = 1 M θ ^ m where M is the number of completed data sets and θ ^ m, m = { 1, …, M } is the estimate of the parameter of interest provided by the m th completed data set.
