One more comment. Data sharing is a mixed blessing, to say the least. While it may prevent the petty fraud of outright misinterpretation of statistical results, it may lead to a much greater problem of propagating biased results. The data are not facts, but rather constructs produced by data collection procedures. Those procedures have their own biases that result in preferring one kind of data over another. This implicit bias is seldom apparent and rarely identifiable in statistical analyses. Consequently, sharing the same biased data will result it propagation of biased results.

The failure to share data forces scientists to produce their own data independently. The meta-analysis of these independently collected data sets cancels out the biases inherent in individual data sets, and thus offers a more objective method of hypothesis testing than repeated analysis of the same data set.