Todo

[dheimgartner]

See: #42

• Add convergence diagnostics (standard second order conditions) to summary.opsr()
  but probably in a separate opsr_diagnostics() function:
  • Maximum absolute gradient.
  • Whether the Hessian is positive definite.
  • Possibly eigenvalues or conditioning diagnostics (?).
• What simulation exercise did Lokshin perform (see literature)?
• Replicate (or extend).
  • What is challenging for identification?
• Implement metrics to provide empirical evidence for identification.
• Rewrite paper.
  • Remove [OK] [WARNING] flags from the diagnostics.
  • Remove --- 1. LR test: ... --- and => --- from print() for exclusion test object.
  • Remove ; in 3.2, 4. (and other sections).
  • Use 1000 replications in the simulation study!
  • Instead of using itemize in the design subsection, use a table.
  • Make table 2 and 3 conform with the other tables (caption below).
  • Add "Run" to the table in the section on the simulation study.
  • Run 3 is omitted from the paper (included in the stored object) to match the numbering of Chiburis and Lokshin (2007). (?)
  • Exclusion restriction is justified twice => remove one.
  • The hessian is not negative definite in the case study! => check specification in issue again.
  • The LR-test in the case study does not reject the null, or does it?
  • Why is Figure 2 truncated? shouln't the curve span [-0.99, 0.99]?
  • Case study needs a rewrite (after the final specification has been identified).
  • Manually restructure the paper => maybe add sample selection comp to appendix?
• Make sure man is in sync.
  • opsr_step() correct p-value above and add hint/warning from vignette.
  • Is the simulation_study dataset documented?
• Update on CRAN.
• Resubmit: When you have uploaded Revisions for all required files above
  (including your point-to-point response), you need to inform the editors that
  your revision is complete. To do so click the Add discussion button below, using
  the subject "Resubmission" and adding all involved section editors and journal
  editors (if any) as participants.