Updated ERF functions and sensitivity analysis

Changes:

  • added RRs of linear estimates per 10 dB from ERF analysis each graph/outcome

  • added RRs for simple meta-analysis to each graph

  • added sensitivity analysis for low exposure bias studies only

This is an extension to previous analysis reported available here https://rpubs.com/andpyk/meta-6-4-Updated-ERF

IHD, MI, and REML vs simple meta-analysis

Linear Model fitted with non-linear input data from all IHD studies is expressed per 10 dB(A) increase in road traffic noise (pink). Additionally, the model is overlayed with the RR and 95% CI from the simple meta-analysis (light blue).

Figure 1: REML Linear Model, IHD, with RR per 10 dB
Figure 2: REML Linear Model, overlayed with IHD RR from simple meta: 1.017 (0.990, 1.044)

Linear Model fitted with non-linear input data from all MI studies is expressed per 10 dB(A) increase in road traffic noise (pink). Additionally, the model is overlayed with the RR and 95% CI from the simple meta-analysis (light blue).

Figure 3: REML Linear Model, MI, with RR per 10 dB
Figure 4: REML Linear Model, MI, overlayed with MI RR from simple meta: 1.029 (1.011, 1.048)

Linear Model fitted with non-linear input data from all Stroke studies is expressed per 10 dB(A) increase in road traffic noise (pink). Additionally, the model is overlayed with the RR and 95% CI from the simple meta-analysis (light blue).

Figure 5: REML Linear Model, STROKE, with RR per 10 dB
Figure 6: REML Linear Model, STROKE, overlayed with STROKE RR from simple meta: 1.025 (1.009, 1.041)

Sensitivity analysis (Low Exposure Bias studies only)

Figure 7: Input non-linear estimates from studies with low exposure bias
Figure 8: REML Spline-based model, IHD (Low Exposure Bias studies only)
Figure 9: MI: Input non-linear estimates from studies with low exposure bias
Figure 10: REML Spline-based model, MI (Low Exposure Bias studies only)
Figure 11: Stroke: Input non-linear estimates from studies with low exposure bias
Figure 12: REML Spline-based model, STROKE (Low Exposure Bias studies only)