Friday, April 04, 2014

Harold Winter on Empirical Study

In practice, empirical analysis can be extremely complicated and contentious. Also, real-world data can be difficult to obtain or inaccurately measured.


Furthermore, there may be alternative ways to measure the same variable, or there may be a question of which data are relevant.


In all, there is room for much legitimate debate in empirical research. In my opinion, the best empirical work deals openly with these shortcomings. Many authors allow their data to be shared by other researchers.

結局、実証研究には多くの合理的な議論の余地があるということだ。 私の見解では、最も素晴らしい実証研究はこれらの欠点を公然と扱っている。多くの著者が自らのデータを他の研究者と共有することができるようにしている。

Being able to have someone else independently replicate your results is the first step toward having confidence in the integrity of your data. The second step is to allow a battery of tests to check for the robustness of your results.  


Finally, again in my opinion, no single empirical study can truly be definitive. It may even be impossible to consider a consensus body of empirical research as definitive.


Winter(2008) pp10  *Italics are original in Winter.

It certainly may be prudent to be suspicious of empirical results that support any preconceived bias the researchers are known to have, but the bias in and of itself does not refute the results.


Any study can be stained with bias. What's most important is to be able to verify the integrity of the data and to be able to replicate the results.


Furthermore, are the results robust to different configurations of the data and different statistical techniques? Simply criticizing or dismissing a study because of its results is not constructive.


Winter(2008) pp113  *Italics are original in Winter.

No comments: