|
Abstract
|
We thank Andrew Conner and colleagues for their careful and thoughtful review of our Article. We greatly appreciate the feedback.
We have a few clarifications on the points they have raised. First, the network meta-regression we used is capable of accounting for consistent differences between case definitions in datasets, allowing us to combine spatial-temporal information from different sources and adjust them to the level of a gold standard. For this reason, The Counted and Fatal Encounters can use different case definitions and both be included in the regression. We excluded vehicle deaths from Fatal Encounters but not from The Counted, our gold standard case definition, because vehicle deaths where police were present but not directly involved are very high in Fatal Encounters, which could have altered the temporal trend of these data in a way that the regression could not account for.
As such, we acknowledge that sources like the National Violent Death Reporting System (NVDRS) did not necessarily need to be excluded on the basis of their case definition alone. However, the NVDRS was also excluded because this study was a sub-national analysis, and the regression approach we used required coverage of all 50 US states and District of Columbia by each dataset to be solvable. The NVDRS did not make the data available for every state. If comprehensive state-level data are available through a request process that we were not aware of, we would be happy to include it in future iterations of police violence estimates. However, we do agree that because the NVDRS is an official surveillance system with notable coverage, we should have discussed it further in the Discussion section rather than in the appendix. Thank you for bringing this gap to our attention.
Second, we understand that each source uses a different case definition, but we believe that the case definition we used is the most accurate to capture police violence. Our definition is used by both Mapping Police Violence and The Counted. However, if we are able to attain NVDRS data from all 50 US states and District of Columbia in the future, we would be happy to include it with its respective case definition in our analyses...
Language: en |