Adding CDS bootstrapping

[farzadwp]

In this commit I have

• Added new classes to contain CDS, survival probabilities, and hazard rates curves;
• Added new classes to contain CDS curve specifications;
• Added generic methods to format and print for the above classes; and
• Created bootsrapping generic methods

All of the new classes and methods contain full documentation with examples
@imanuelcostigan

[imanuelcostigan]

Questions:

1. Is there a nice way that you can combine the HazardCurve and SurvivalCurve classes into one? Looks like the only reason for separating them is because credule::bootstrapCDS() returns hazard rates and survival probabilities in one object and you want to distinguish them. Possibilities:
   • can you extend CDSCurve by including hazard_rates and survival_probabilities fields which are set to NA_numeric if the instance is not bootstrapped and otherwise the bootstrapped values? OR
   • define a class called BoostrappedCDSCurve which includes hazard_rates and survival_probabilities fields (plus others) which is returned by bootstrap() where the latter is a generic that replaces bootstrap_survprob() and bootstrap_harzardrate().
2. Where do the valid rank labels come from? Would be better if we could make these nicer labels that have a consistent convention. All lower case would be better (the user doesn't have to remember). For eg. Senior -> "senior", SubTier3 -> "sub_tier_3", SubTier2Upper -> "sub_tier_2_upper", SubTier2Lower -> "sub_tier_2_lower", SubTier1 -> "sub_tier_1"

Actions:

• I have added a few commits to yours on the imanuelcostigan:feature-cds-ic branch. (Long story about why I did this). Can you merge these into your fork? Happy to show you how to do this.
• Please review my commit commentary to understand my changes.
• Update your version of roxygen2
• Use the print / format pattern of the ZeroCurve class for the various credit curve classes you have introduced (main thing is that it goes via a tibble).
• Document whether the probabilities passed to SurvivalProbCurve are cumulative or marginal survival probabilities
• Review the CONTRIBUTING file on the repo as it relates to PR and make sure you have added test cases, updated the NEWS file, etc
• Update the README.Rmd (and knit the corresponding README.md file)
• Create a vignette for the credit curve functionality
• Ensure there are no errors, warnings or notes returning from devtools::check()

[farzadwp]

1- In this iteration, the new credit classes mirror interest rate classes in functionality, as follows:
SurvivalProbabilities <-> DiscountFactors
ZeroHazardRate <-> InterestRate
CreditCurve <-> ZeroCurve

2- I have limited the use of Credule package (for bootstrapping) to only one generic: as_SurvivalProbabilities.CDSCurve

3- CreditCurve constructor only needs Survival Probabilities, so it doesn't have any dependency on the bootstrapping. I chose the name CreditCurve instead of the suggested BootstarppedCDSCurve, because the class no longer depends on bootstrapping.

4- Zero Hazard rates are now calculated from Survival Probabilities, without having to do an additional bootstrapping. (This is to ensure internal consistency in terms of business conventions)

5- I have extended all of the interpolation generic functions, to mimic similar functionalities for the credit classes. Here, note that
I overload interpolate_dfs with CreditCurve, instead of defining a new generic (e.g. interpolate_sps)

6- There are additional rules for CDSSpecs in cred-ops. When output specs are ambiguous, they are replaced with an "Empty" CDSSpecs object.

7- for +,-,/ and *, I included an assertion to check if the length of objects are equal. In IR and DF operators, if the length are not equal, equally lengthed vectors are created. I suggest including an assertion for IR and DF operators as well.

---

Point-to-point responses to comments:

Question 1- In this iteration, Credit Curve holds that responsibility, in the same way that ZeroCurve does for rates.
Question 2- Discussed in the meeting
Action 1 - Done
Action 2- Done
Action 3- I updated my Roxygen2 to 6.1.0
Action 4- Done
Action 5- I added this to as_SurvivalProbabilities.CDSCurve
Action 6- I added some tests on the important functions. Note that context() is being deprecated, so I did not use it.
Action 7- Done
Action 8- Check() fails on running fmbasics-Ex.R, which I did not edit. I haven't been able to figure out the cause of the issue.

[imanuelcostigan]

Please:

1. refresh your PR from master and resolve conflicts
2. make sure all checks pass with no warnings, notes etc

[farzadwp]

Please review the pull request. I have
1- refreshed your PR from master and resolve conflicts
2- resolved errors, warnings, and notes