Changes in version 3.1.4 (2025-02-14)                  

Changes

1.  Removing extreme estimates (relative risk < 0.01 or relative risk
    > 100) before fitting null distribution.

Bugfixes

1.  Changed initial parameter values when fitting null distribution
    using asymptotics and MCMC, avoid errors in some edge cases.

                 Changes in version 3.1.3 (2024-09-30)                  

Changes

1.  fitNull() and fitMcmcNull(), as well as downstream calibration
    functions, throw no error when there are no valid negative controls
    estimates (instead returning NAs).

                 Changes in version 3.1.2 (2023-12-21)                  

Bugfixes

1.  Allowing fitNull() to work with many estimates (n > 1000).

2.  Fixed edge case where fitMcmcNull() would get stuck in an infinite
    loop.

                 Changes in version 3.1.1 (2022-08-09)                  

Changes

1.  Making sure we pass R check even if suggested packages are
    unavailable, as required by CRAN.

                 Changes in version 3.1.0 (2022-07-01)                  

Changes

1.  Adding the compareEase() function to compare EASE of correlated sets
    of estimates.

2.  Adding the computeCvPoissonRegression() function to compute critical
    values for Poisson regressions.

3.  The plotCalibrationEffect() function warns if there are values
    outside the plot limits.

4.  Changing the MaxSPRT calibration vignette to use CV calibration
    instead of LLR calibration.

Bugfixes

1.  Fixing computeTraditionalP() when using vectors as input.

2.  Fix error when no critical value meets requirements (throwing
    warning instead).

                 Changes in version 3.0.0 (2021-09-28)                  

Changes

1.  Adding option to make p-values one-sided.

2.  Adding empirical calibration of the log likelihood ratio.

3.  Adding functions for computing critical values for the log
    likelihood ratio when performing sequential testing (MaxSPRT).

4.  Adding ability to fit null distribution using non-normal
    approximations of the per-negative control likelihood functions.

5.  Faster null distribution fitting using MCMC. Increasing default MCMC
    iterations for greater stability.

6.  Using median instead of mean of posterior distributions when
    converting MCMC null to error model for greater consistency between
    calibrated CI and P.

                 Changes in version 2.1.0 (2021-03-04)                  

Changes

1.  Adding computation of expected absolute systematic error. Can be
    shown in plots.

                 Changes in version 2.0.2 (2020-04-07)                  

Changes

1.  Center title on all plots.

Bugfixes

1.  Fix build fail in R 4.0.0.

                 Changes in version 2.0.1 (2020-01-13)                  

Changes

1.  computeTraditionalCi now outputs data frame instead of vector.

Bugfixes

1.  convertNullToErrorModel function now adheres to new systematic error
    model (SD no longer on log scale).

                 Changes in version 2.0.0 (2019-07-08)                  

Changes

1.  Adding new method for fitting systematic error models that drops the
    assumption that SD is linear in the log scale, but rather just
    linear (in the logRr space). Setting that to default, but allowing
    users to use legacy model if they choose.

                        Changes in version 1.4.1                        

Changes

1.  Added convertNullToErrorModel function to allow confidence interval
    calibration using only negative controls (requiring the user to make
    some assumptions).

                 Changes in version 1.4.0 (2018-11-24)                  

Changes

1.  Added plot showing effect of confidence interval calibration,
    similar to p-value plot.

Bugfixes

2.  Fixed 'unknown aesthetics' warning when calling plotTrueAndObserved.

                 Changes in version 1.3.6 (2017-11-07)                  

Changes

1.  Added plots showing expected type 1 error given an estimated
    empirical null distribution.

2.  Closed form solution for RR vs SE plot for faster computation
    (especially when plotting credible intervals).

Bugfixes

1.  Several improvements of the robustness of fitting systematic error
    models.

                 Changes in version 1.3.1 (2017-05-16)                  

Changes

1.  Confidence interval calibration model StdDev transformed to log
    scale to prevent negative StdDev.

2.  Confidence interval calibration cross-validation now allows
    specification of leave-out groups.

3.  various new plots for evaluating confidence interval calibration.

4.  Added confidence interval calibration vignette.

5.  Added example data for confidence interval calibration

Bugfixes

1.  None

                 Changes in version 1.2.0 (2016-08-16)                  

Changes

1.  Ability to add credible intervals to calibration effect plot

2.  Plot CI calibration (using leave-one-out cross-validation)

Bugfixes

1.  Fixed vignette name in index

2.  Removed coverage plot (moved to MethodEvaluation package)

                 Changes in version 1.1.0 (2016-02-15)                  

Initial submission to CRAN