Table 3 Methodology characteristics in included ITS

Description of the analysis

115 (99)

 Segmented regression

90 (78)

 ARIMA model

15 (13)

 Generalised estimating equations

7 (6)

 Change-point analysis

2 (2)

 Mixed model

1 (1)

Autocorrelation was considered

63/115 (55)

Method used to test for autocorrelation^a

N = 63

 Durbin Watson

22 (35)

 Autocorrelation function

13 (21)

 Partial autocorrelation function

11 (17)

 Ljung-Box

3 (5)

 Examination of residuals

2 (3)

 Portmanteau tests

2 (3)

 Autocorrelation probability

1 (2)

 No test performed

23 (37)

Autocorrelation present if a test was performed

N = 40

 No

12 (30)

 Yes

25 (63)

 Not stated

3 (8)

Method used to adjust for autocorrelation

N = 48

 Autoregressive error term

14 (29)

 Prais-winsten regression model

8 (17)

 Differencing

3 (6)

 Newey-west standard errors

2 (4)

 Yule-walker regression model

2 (4)

 Did not specify

19 (40)

Order of autocorrelation

N = 48

 1

8 (17)

 2

3 (6)

 3

2 (4)

 4

1 (2)

 5

1 (2)

 Did not specify

33 (69)

Nonstationary was considered

9/115 (8)

Method used to test for non-stationary

N = 9

 Dicky-Fuller

5 (56)

 ACF and PACF

2 (22)

 Significance testing

1 (11)

 Not stated

1 (11)

Nonstationary was present if a test was performed

N = 8

 No

4 (50)

 Yes

3 (38)

 Not stated

1 (13)

Method used to adjust for nonstationary

N = 3

 Differencing

2 (67)

 Within the ARIMA model

1 (33)

Seasonality was considered

28/115 (24)

Method used to test for seasonality

N = 28

 ACF PACF

5 (18)

 Regression diagnostic tests

1 (4)

 Dicky-Fuller

1 (4)

 Just stated a test was performed

4 (14)

 No formal test

17 (61)

Seasonality present if a test was performed

N = 7

 No

6 (86)

 Yes

1 (14)

Method used to adjust for seasonality

N = 22

 Covariate

7 (32)

 Seasonal ARIMA

1 (5)

 Differencing

1 (5)

 Not stated

13 (59)

Sample size description

N=115

 No

108 (94)

 Yes

7 (6)