Skip to main content

Table 1 Results from the methodological evaluation of the articles in this review

From: Poor quality in the reporting and use of statistical methods in public health – the case of unemployment and health

Criterion

Number

Percent

Events per variable/overfitting (n = 40)

  

 Fulfilled

39

97.5 %

 Not fulfilled

1

2.5 %

Conformity with a linear gradient (n = 19)

  

 Fulfilled

0

0 %

 Not fulfilled

19

100 %

 No continuous variable in multivariate analysis

22

 

Interactions between independent variables (n = 41)

  

 Interaction terms used

14

34 %

  Coefficients tested

14

 

  Coefficients not tested

0

 

 Stratified results presented

27

66 %

  Test for interactions

3

 

  No test for interactions

24

 

 Neither stratified results nor interaction terms discussed

7

17 %

Collinearity of independent variables (n = 39)

  

 Discussed

1

2.6 %

 Not discussed

38

97.4 %

Validation of the statistical model (n = 41)

  

 Fulfilled

0

0 %

 Sensitivity analysis/robustness checks

3

7.3 %

 Not used

38

92.7 %

Statistical significance/confidence intervals (n = 41)

  

 Statistical significance for statistical model

2

5.1 %

 No statistical significance for statistical model

39

94.9 %

P-value only presented as above/below the significance level

14

34 %

 Confidence interval presented for exposure

24

58 %

Coefficients presented in statistical models with relative difference (n = 23)

  

 All coefficients presented for the statistical model

13

52 %

 Not all coefficients presented for the statistical model

12

48 %

Goodness of fit test (n = 41)

  

 Fulfilled

9

22 %

 Not fulfilled

32

78 %

Selection of variables (n = 41)

  

 Selection of candidate variables discussed

13

32 %

 Selection of candidate variables not discussed

28

68 %

 Selection of final variables in statistical model discussed

5

12 %

 Selection of final variables in statistical model not discussed

36

88 %

Coding of variables (n = 41)

  

 No faults

27

66 %

 Small faults

11

27 %

 Major faults

3

7.3 %

Fitting procedure (n = 41)

  

 Performed

3

7.3 %

 Not performed

38

92.7 %

Discussion of methodological concerns (n = 41)

  

 Yes

31

76 %

 Limitations with statistical model

6

15 %

 No

10

24 %