Figure 3.

Effect of different delays in data availability on model quality. As the delay in availability of data on suicide occurrences varies for the public and policymakers, we perform the analysis for a range of possible delays. Specifically, we re-estimate the Google model in Eq. (3) using lags of 1, 3, 6, 9, 12, and 18 months, in addition to the original reporting delay of 24 months (x-axis). This covers delays in data availability for public (24 months delay), policymakers (around 6 months delay) and hypothetical models (a delay below 6 months). The adjusted $R^2$ values for the base model (dashed blue line) and the Google model (solid red line) with corresponding delays are shown on the y-axis. We observe that the results do not vary greatly when the lag in data availability is modified. The only mild deviation is observed for hypothetical delays of one or three months, where the base model improves, but in these cases too, the Google model still performs better than the base model. This provides further evidence of the potential value of online data in estimating suicide occurrence statistics.