AIR QUALITY CONTROL USING T ² HOTELLING DOUBLE BOOTSTRAP CONTROL CHART WITH VECTOR AUTOREGRESSIVE MODEL APPROACH

This study aims to find a sensitive control chart to shifts in the control process that induces a correlation among time series observations, known as autocorrelation. T ² Hotelling, a popular multivariate chart, is no longer sensitive to detect small and moderate mean shifts derived from the autocorrelation process. Therefore, this study uses Vector Autoregressive Model (VAR) residuals to build T ² Hotelling control charts. To improve the chart Double Bootstrap method is used to construct a sensitive control limit because the assumptions of the T ² Hotelling are not fulfilled. Violation of assumptions results in the analysis being inappropriate. The proposed control chart is used for air quality control in Surabaya with the characteristics quality of PM 2.5, PM 10, and CO, which are correlated with each other. The proposed control chart’s performance is compared with the single Bootstrap control chart by Average Run Length (ARL) value at different numbers of observations. The results show that The Proposed T ² based on residual VAR with Double Bootstrap is more sensitive than the single Bootstrap to detect out-of-control on all shifts and at observations. Thus, the proposed control chart can be a way to minimize errors in controlling air quality.