Skip to main content

Table 2 Estimated regression coefficients and standard errors (SE) when no values are missing; when data are missing completely at random; when outcome blood pressure (BP) is missing at random; when covariate (baseline BP) is missing at random; and when outcome BP is missing not at random. When values were missing, multiple imputation as well as the maximum likelihood method were used

From: When and how should multiple imputation be used for handling missing data in randomised clinical trials – a practical guide with flowcharts

Type of missingness

Analysis

Regression coefficients

Intercept

estimate

(standard error (SE))

P

Baseline blood pressure

estimate

(SE)

P

Outcome blood pressure

estimate

(SE)

P

No missing values

Complete case analysis

−2.48

(4.69)

0.60

1.013

(0.025)

<0.0001

−50.8

(1.48)

P < 0.0001

Missing completely at random (MCAR)

Multiple imputation

−6.11

(5.72)

P = 0.29

1.037

(0.030)

P < 0.0001

−51.5

(1.78)

P < 0.0001

Maximum likelihood

−6.85

(5.17)

P = 0.18

1.041

(0.028)

P < 0.0001

−51.2

(1.68)

P < 0.0001

Missing at random (MAR)

(outcome)

Multiple imputation

−2.60

(5.15)

P = 0.61

1.014

(0.028)

P < 0.0001

−51.0

(1.70)

P < 0.0001

Maximum likelihood

−2.75

(5.08)

P = 0.59

1.015

(0.027)

P < 0.0001

−51.2

(1.65)

P < 0.0001

Missing at random (MAR)

(baseline blood pressure)

Multiple imputation

−6.09

(5.37)

P = 0.26

1.026

(0.029)

P < 0.0001

−51.1

(2.16)

P < 0.0001

Maximum likelihood

−5.49

(5.41)

P = 0.31

1.026

(0.032)

P < 0.0001

−50.2

(2.18)

P < 0.0001

Not missing at random (MNAR)

(outcome blood pressure)

Multiple imputation

−8.64

(5.07)

P = 0.089

1.026

(0.028)

P < 0.0001

−47.5

(1.99)

P < 0.0001

Maximum likelihood

−8.13

(5.61)

P = 0.15

1.026

(0.032)

P < 0.0001

−47.6

(2.09)

P < 0.0001

  1. For comparison the results of an analysis of the data without any values missing is also shown