However, we imputed those BMI with missing values, based on a regression imputation model including all risk factors such as age, sex, smoking status, blood pressure, cholesterol level and pension in the model.
We conducted tests for trend by including the continuous variable (values below the DL were based on one imputation) in the regression models.
Table 3 shows that estimates of μ based on Tobit regression, multiple imputation, and single impute fill-in approaches are generally unbiased.
The imputation was based on Bayesian regression that imputes missing observations for every patient using all available data (i.e., all repeated measures on weight up to the patient's discontinuation from the study).
Single imputation, such as mean imputation or imputation based on linear regression, leads to incorrect statistical tests because the complete-data analysis does not account for uncertainty created by the fact that data are missing [ 5, 6].
Multiple imputation is also employed to assess the uncertainty due to imputation since results based on imputation alone, especially when linking imputed locations to other spatial data, may be questionable.
For the moderately skewed measure, the C-GHQ, imputation based on linear regression with no rounding had the least biased point estimate under both the MCAR and MAR scenarios.
A common approach to imputation is based on linear regression analysis.
Overall, the imputation scenario based on polytomous regression seems to yield the best results.
The imputation models based on logistic regression and predictive mean matching included the interaction between survival time and censoring in their set of predictor variables, because a short survival time must be interpreted differently for a deceased person than for someone still alive, having only a short follow-up time.
