Skip to main content
Fig. 4 | BMC Medical Research Methodology

Fig. 4

From: Comparison of statistical methods used to meta-analyse results from interrupted time series studies: an empirical study

Fig. 4

Bland Altman plot of difference in standardised meta-analytic effect estimates (y-axis) vs average of the effect estimates (x-axis), for each pairwise comparisons of ITS analysis and meta-analysis method combination (top row of the label indicates the ITS analysis methods, bottom row indicates the meta-analysis method, e.g., OLS ITS DL MA is OLS ITS analysis with DerSimonian and Laird between-study variance meta-analysis). The top triangle (green points) presents the immediate level-change (difference calculated as column method – row method), and the bottom triangle (blue points) presents the slope-change per month (difference calculated as row method – column method). Horizontal orange lines depict the average, dashed orange lines depict the 95% limits of agreement (calculated as the mean ± 1.96*standard deviation of the differences). Vertical grey line indicates an average of zero, while the horizontal grey line indicates a mean difference of zero. The coloured boxes indicate cells that compare ITS analysis methods when fixed-effect meta-analysis was fitted (red boxes), meta-analysis models (i.e., fixed- vs random-effects models)[yellow boxes], between-study variance estimators (i.e., using DL or REML)[blue boxes], and combinations of between-study variance estimators and ITS analysis methods (black boxes). The solid coloured boxes indicate comparisons of level-change and dashed boxes indicate slope-change per month. DL DerSimonian and Laird, HKSJ Hartung-Knapp/Sidik-Jonkman, ITS interrupted time series, MA meta-analysis, OLS ordinary least squares, REML restricted maximum likelihood, WT Wald-type

Back to article page