Open ReidConor opened 7 years ago
ReidConor
commented
7 years ago See papers in #4
Going with
ReidConor
commented
7 years ago Feedback to action on:
[x] "the exploration of causal research" -> causality research
[x] "It has been demonstrated to be quite reliable in a variety of contexts and countries." -> probably should be quoted inline rather than a block quote
[x] "Figure 2.1: Correlation between corporate environmental protection spending and economic success." -- put the image source in the caption.
[x] ”the responsibility of enterprises for their impacts on society” -- here and elsewhere, you need to use latex-style quotes, that is double backticks (``) followed by double apostrophes ('').
[x] "For example, the study of Moldovan and Mutu (2015) on which the current study is based, makes strong claims as to the relationship between corporate governance and company success." -- here, I would quote something from their conclusions where they suggest that firms could improve their Z score if they made some specific change.
[x] "It is said that if all variables that could possibly be causal are considered, causation can be reliably inferred." -- I don't quite get this comment and "it is said" needs to be backed up.
[x] "Researchers can then estimate the presence and magnitude of the effect the treatment has on particular outcomes." -- I would add a sentence here to make it crystal-clear: any observed relationship can then be inferred to be causal since the treatment variable was under researcher control.
[ ] "A way to address this is to categorise continuous features, a practice used in calculating the Mahalanobis distance for example which works well with low dimensionality, but poorly with highly non-uniformly distributed features." -- I'm not familiar with this -- I think of Mahalanobis as working on continuous features -- so just flagging it for future discussion. (Or maybe you have a cite.)
[x] I suggest a find-replace on your .tex for "casual" -> "causal" ;-)
[ ] How does the DAG model of Pearl hold up when we think about variables which are in a circular i.e. feedback relationship, e.g. sales go up -> more money available to spend on advertising -> causes sales to go up? Pearl has talked a bit about this. I think you don't need to go into it, but suggest add a few sentences saying that it is possible, but more complicated, and not necessary for our project.
[ ] More broadly, I think there is another approach which is analogous to Matching, that is a before-and-after matching on the same data point. Eg if you could get the Altman Z for each company before and after each change to its governance (how far before and after? difficult question) you would in a sense have a matched dataset. Again, just something to mention. Might not be needed for us.
ReidConor
commented
7 years ago "It is said that if all variables that could possibly be causal are considered, causation can be reliably inferred." -- I don't quite get this comment and "it is said" needs to be backed up.
I've changed this to read "Logically, one must consider in their analysis all variables that could possibly prove to be casual, least a conclusion be drawn that leaves out some feature that in fact is the underlying casual factor for some observed phenomenon. Of course in practical studies this is almost never fully achieved due to real-world complexity, and so it may be that advanced techniques are required to move towards this case as much as possible to overcome this issue."
See feedback and adjust Lit Review accordingly.