Skip to main content

Table 3 Models imputing outcome variables

From: Heckman-type selection models to obtain unbiased estimates with missing measures outcome: theoretical considerations and an application to missing birth weight data

Outcome variable

Birth weight in grams

Variable

Imputation

PIDA

Mean imputation

MICE

Group Mean replacementa

-0.5SDb

+ 0.5SDc

 IFAS: No

 Yes

7.02 (−13.97; 28.01)

22.38 (−13.49; 58.26)

18.71* (−2.27; 39.69)

−51.78*** (−73.76; − 29.80)

76.78*** (54.99; 98.57)

Educational attainment: No schooling(+)

 Primary

10.96 (−9.75; 31.66)

20.03 (− 12.71; 52.77)

10.98 (−9.72; 31.68)

13.73 (− 7.96; 35.41)

7.98 (− 13.52; 29.48)

 Coranic

22.96 (− 19.61; 65.52)

36.12 (− 26.93; 99.17)

23.34 (−19.22; 65.89)

36.71 (− 7.87; 81.29)

8.31 (−35.88; 52.51)

 Secondary or higher

32.13* (−0.03; 64.30)

49.47** (0.90; 98.03)

32.09* (− 0.06; 64.25)

45.41*** (11.72; 79.09)

17.79 (− 15.61; 51.18)

Maternal age (years): 20–34(+)

 15–19

− 133.30*** (− 160.60; − 106.00)

−209.30*** (− 253.10; − 165.50)

−133.92*** (− 161.23; − 106.61)

−131.68*** (− 160.29; −103.07)

− 135.34*** (− 163.71; − 106.98)

 35–49

9.22 (− 19.01; 37.44)

6.91 (− 36.48; 50.30)

9.46 (−18.76;- 37.67)

24.99* (− 4.57; 54.54)

− 7.68 (− 36.98; 21.62)

Socioeconomic status: Most poor(+)

 Poor

14.68 (−14.16; 43.52)

32.48 (− 18.42; 83.39)

14.43 (− 14.40; 43.27)

21.87 (−8.33; 52.07)

6.79 (− 23.16; 36.74)

 Middle

13.04 (− 17.00; 43.08)

25.15 (− 24.63; 74.93)

12.77 (− 17.26; 42.79)

42.02*** (10.57; 73.48)

− 18.40 (− 49.58; 12.79)

 Rich

−1.75 (− 31.22; 27.71)

− 0.52 (− 53.46; 52.41)

−1.93 (− 31.39; 27.52)

27.05* (− 3.81; 57.91)

− 32.94** (− 63.54; -2.35)

 Most rich

35.31* (− 0.68; 71.30)

53.72* (− 4.44; 111.90)

35.06* (− 0.92; 71.03)

66.49*** (28.80; 104.18)

1.52 (− 35.84; 38.89)

Child sex: Male(+)

 Female

− 80.03*** (− 97.96; − 62.09)

− 125.20*** (− 153.80; − 96.59)

− 80.17*** (− 98.10; − 62.24)

− 80.28*** (− 99.06; − 61.50)

−79.83*** (− 98.45; − 61.21)

Previous births: 0 child(+)

 1–4

44.13* (− 2.89; 91.15)

57.12 (− 17.29; 131.50)

44.35* (− 2.65;- 91.35)

40.49 (− 8.75; 89.73)

48.18* (− 0.64; 96.10)

 ≥5

112.30*** (59.65; 164.90)

186.30*** (100.60;- 272.00)

112.38*** (59.77; 165.00)

94.30*** (39.17; 149.43)

131.75*** (77.10; 186.40)

Twin births: No(+)

Yes

− 476.80*** (−518.80; − 434.80)

− 642.10*** (− 702.90; − 581.40)

−476.64*** (− 518.62; − 434.66)

− 452.14*** (39.18; 149.42)

− 503.37*** (− 546.98; − 459.78)

observations

7325

7325

7325

7325

7325

  1. MICE Multiple imputations by chained equation. MICE was done with 150 imputations using Stata mi estimate package
  2. PIDA Pattern imputation with delta adjustment
  3. aMissing values were replaced with mean of group (mean of observed birth weight for treated, and mean of observed birth weight for non-treated for control)
  4. bMissing values were replaced with a birth weight half a standard deviation lower than the observed mean.
  5. cMissing values were replaced with a birth weight half a standard deviation above the observed mean.
  6. *** p < 0.01. ** p < 0.05.* p < 0.1;(+) Reference category; Adjusted model controls village fixed effects; CI: confidence interval
Back to article page

BMC Medical Research Methodology

ISSN: 1471-2288

Contact us