Skip to main content

Table 2 Time series model selection criteria for TB risk residuals

From: Time series analysis of demographic and temporal trends of tuberculosis in Singapore

Ā 

Lag 0

Lag 4

Model (Non-residents)

AIC

MSPE

AIC

MSPE

AR(2)

āˆ’42.79

5.01

āˆ’37.6

5.13

MA(2)

āˆ’42.88

5.06

āˆ’36.39

5.09

ARMA(2,2)

āˆ’46.08

4.5

āˆ’35.42

4.8

SARIMA(1,0,0)(2,0,0)[14]*

āˆ’51.1#

4.54#

āˆ’46.49#

4.28#

Model (Residents)

AIC

MSPE

AIC

MSPE

AR(2)

āˆ’42.66

0.85

āˆ’51.67

1.56

MA(3)

āˆ’39.05

0.86

āˆ’51.35

1.57

ARMA(2,2)

āˆ’39.17

0.87

āˆ’48.36

1.57

SARIMA(1,0,0)(2,0,0)[14]*

āˆ’49.18#

0.77#

āˆ’59.42#

1.48#

  1. AIC; Akaike information criterion, MSPE; Mean squared prediction error, AR; Autoregressive, MA; Moving average, ARMA; Autoregressive moving average, SARIMA; Seasonal Autoregressive Integrated Moving Average,
  2. *; A SARIMA model for residual, with 12Ā month seasonality, seasonal autoregressive component of order 2 and linear autoregressive component of order 1.
  3. #; lower AIC and MSPE values indicate better fit of the model.
Back to article page

BMC Public Health

ISSN: 1471-2458

Contact us