Skip to main content

Advertisement

Springer Nature is making SARS-CoV-2 and COVID-19 research free. View research | View latest news | Sign up for updates

Table 2 Specification of the missing data mechanisms to be imposed

From: Comparison of techniques for handling missing covariate data within prognostic modelling studies: a simulation study

Mechanism X3 (PGR) X2 (LN) X5 (TRT) X8 (TS)
MCAR β0 β0 + ln(OR)MX3 β0 + ln(OR)MX2 β0 + ln(OR)MX3
MAR ln(0.8)X4 ln(3)X1 ln(0.7)ln(t) ln(7)X7
MNAR ln(1.3)X3 ln(0.6) X2 ln(8)X5 ln(0.9)X8
COMBINED   ln(0.7)ln(t) +
ln(0.3)X5
ln(3)X1 ln(0.9)X8
  1. Note: A logistic regression model was used to model the probability of missingness for each incomplete covariate. The entries in the table represent the variables associated with the missingness of each incomplete covariate. For MAR, MNAR, and the combined mechanism, the terms given are extra to those for the MCAR mechanism, e.g. the MAR mechanism for X2 is
  2. where β0 is the intercept, estimated by solving the above equation using the specified probabilities of missingness for X2 and X3 and the average covariate value of X1, MX3 is the missingness indicator for covariate X3, which equals 1 if an observation is missing and 0 if the value is observed and OR is odds ratio for the relationship between the missingness of X2 and X3, and is obtained from Table 3. The coefficients for the variable associated with the mechanism were modified from relationships with missing data seen in another study [27] to provide significant associations. All continuous variables including survival (t) were standardised by dividing by the standard deviation. When the mechanisms included other covariates subjected to missingness, the original complete data were used.