ssdtools 2.4.0

Features

• Added right censoring for all distributions (by @eduardszoecs).
• Added rescale = "odds" option to ssd_fit_dists() to allow fitting to dilution data including odds_max = 0.999 argument to handle values of 1.

Functions

• Added ssd_ci_methods() and ssd_est_methods() to get character vector of methods.
• Added ssd_at_boundary() and ssd_computable().
• Added class and attribute preserving [[ and [ operators.

Arguments

• Added strict = TRUE argument to subset() to allow subsetting when distributions missing (with strict = FALSE).

Output Columns

• Added column "dists" to ssd_hc(), ssd_hp() and predict() which is a list of the distributions.
• Added "est_method" and "ci_method" columns to tibble output by ssd_hc(), ssd_hp() and predict().
• Added "level" column to tibble output by ssd_hc(), ssd_hp() and predict().
• Added "at_bound" and "computable" columns to tibble output by ssd_gof().
• Renamed "method" column to "boot_method" in tibble output by ssd_hc(), ssd_hp() and predict().
• ssd_hc(), ssd_hp() and predict() now only return missing values for se, lcl and ucl when nboot = 0.

Error Checking

• Added ... and check unused to ensure matching names of subsequent arguments.
• Fixed error message when Inf weights.
• ssd_fit_bcanz() now checks all dists are subset of ssd_dists_bcanz().

Modifications

• Modified method to calculate ssd_ecd() to use stats::ppoints().
• Ensures weighted_samples sum to nboot.

Deprecated

• Deprecated ssd_dists_shiny().

• Deprecated ties.method = "first" argument in ssd_ecd().

• Soft-deprecated est_method = "multi" argument to multi_est = TRUE for ssd_hc(), ssd_hp() and predict().

• Soft-deprecated proportion = FALSE to proportion = TRUE argument to ssd_hp() which switches output from percentages to proportions.

• Soft-deprecated wt = FALSE for wt = TRUE argument to ssd_gof() and glance() which replaces column "weight" with "wt" in output.

• Soft-deprecated ci_method = "weighted_arithmetic" for ci_method = "MACL".

ssdtools 2.2.0

• Added ssd_label_comma_hc() function to label numbers with significant digits and comma and offset hazard concentration value if present in breaks.
• Added ssd_dists_shiny() to define distributions for use in shiny app.
• For plot_coord_scale() function:
  • Added x_limits = NULL to allow setting of x-axis limits.
  • Added hc_value = NULL to pass hazard concentration value.
  • Restricted trans argument to "identity", "log10" or "log".
• For ssd_plot() function:
  • Added text_size argument.
  • Added theme_classic = FALSE argument to switch classic theme.
  • Soft-deprecated size argument for label_size.
• Turned off x-axis minor breaks for all plots (for consistency) as HC major break causing multiple minor breaks in ssd_plot().

ssdtools v2.1.0

• Added ssd_xxmulti_fitdists() functions to accept object of class fitdists.
• Set ssd_xxmulti(..., lnorm.weight = 0) (instead of 1) to avoid incorrect values with do.call("ssd_xxmulti", c(..., estimates(fits)) if fits does not include the log-normal distribution.

ssdtools v2.0.0

ssdtools v2.0.0, which now includes David Fox and Rebecca Fisher as co-authors, is the second major release of ssdtools.

Major Changes

The following changes are major in the sense that they could alter previous hazard concentrations or break code.

Model Fitting and Averaging

Modifications

The following arguments were added to ssd_hc() and ssd_hp()

• multi_est = TRUE to calculate model averaged estimates treating the distributions as constituting a single mixture distribution (previously it was effectively FALSE).
• method_ci = "weighted_samples" to specify whether to use "weighted_samples", "weighted_arithmetic", "multi_free" or "multi_fixed" methods to generate confidence intervals (previously it was effectively "weighted_arithmetic").

In addition the data frame returned by ssd_hc() and predict() now includes a column proportion with values between 0 and 1 as opposed to a column percentage with between 0 and 100.

Finally, with censored data confidence intervals can now only be estimated by non-parametric bootstrapping as the methods of parametrically bootstrapping censored data require review.

Minor Changes

The remaining changes are minor.

Model Fitting

Modifications

The following arguments of ssd_fit_dists() were changed to reduce the chances of the lnorm_lnorm bimodal distribution being dropped from the default set:

• min_pmix = ssd_min_pmix(nrow(data)) so that by default min_pmix is 0.1 or 3/nrow(data) if greater.
• at_boundary_ok = TRUE.
• computable = TRUE.

These changes also allowed the min_pboot = 0.95 argument to be changed from 0.80 for all bootstrapping functions.

It is worth noting that the second two changes also reduce the chances of the BurrIII distribution being dropped.

In addition rescale = TRUE now divides by the geometric mean of the minimum and maximum positive finite values as opposed to dividing by the geometric mean of the maximum finite value to improve the chances of convergence although ssd_fit_bcanz() no longer rescales by default.

Other minor modifications to the model fitting functions include

• estimates.fitdists() now includes weights in returned parameters as well as an all_estimates = FALSE argument to allow parameter values for all implemented distributions to be included.
• delta = 7 instead of delta = 9.21 to ensure weight of included models at least 0.01.
• seeds are now allocated to bootstrap samples as opposed to distributions (which results in a speed gain when there are more cores than the number of distributions).
• lnorm and gompertz initial values are offset from their maximum likelihood estimates to avoid errors in optim().

The following functions and arguments were also added:

• ssd_hp_bcanz() and ssd_hp.fitburrlioz() to get hazard proportions.
• ssd_pmulti(), ssd_qmulti() and ssd_rmulti() for combined mixture distributions.
• ssd_exx() functions to get default parameter estimates for distributions.
• ssd_censor_data() to censor data.
• npars = c(2L, 5L) argument to ssd_dists_bcanz() to specify the number of parameters.
• dists = ssd_dists_bcanz() to ssd_fit_bcanz() to allow other packages to modify.
• samples = FALSE to ssd_hc() and ssd_hp() include bootstrap samples as list of numeric vector(s).
• save_to = NULL to ssd_hc() and ssd_hp() to specify a directory in which to save the bootstrap datasets as csv files and parameter estimates as .rds files.

Fixes

• ssd_hc() and ssd_hp() now return data frame with parametric column.
• ssd_hp() now return data frame with wt column.

Deprecations

The following functions and arguments were deprecated:

• ssd_wqg_bc() and ssd_wqg_burrlioz() were deprecated.
• percent = 5 in ssd_hc() and predict() was soft-deprecated for proportion = 0.05.
• is_censored() is now defunct.

Plotting

Perhaps the biggest plotting change is that ssd_plot_cdf() now plots the average SSD together with the individual distributions if average = NA.

In addition, the following functions and arguments were added.

• scale_fill_ssd() for color-blind fill scale.
• ssd_label_comma() for formatting of x-axis labels.
• trans = "log10" and add_x = 0 to ssd_plot() and ssd_plot_data() to control x-axis scale.
• big.mark = "," for x-axis labels and suffix = "%" for y-axis labels to all plotting functions.

and the following functions deprecated

• comma_signif() was soft-deprecated.
• is_censored(), plot.fitdists(), ssd_plot_cf() geom_ssd() and stat_ssd() are now defunct.

Data

The following data sets were removed

• ccme_data and ccme_boron (available in ssddata package).
• pearson1000 data set.

v1.0.3: ssdtools 1.0.3

- Replaced `size = 0.5` with `linewidth = 0.5` in `geom_hcintersect()` and `geom_xribbon()`.
- Replaced `aes_string()` with `aes()` in examples (and internally).
- Removed use of `tidyverse` package.
- Now tests values to 12 significant digits.
- Fixed description of `ssd_hp()` to be percent affected rather than percent protected.

v1.0.2: ssdtools 1.0.2

- Fixed bug that was producing estimates of 0 for lower HCx values for log-normal mixture model with rescaled data spanning many orders of magnitude.

v0.3.2

add lifecycle to wordlist

v0.3.1

cran comments

v0.2.0

update cran comments

v0.1.0

update docs