Improve validation

This release includes several major changes related to reading, checking, and validating resistance outcome data:

• [Breaking Change] the debark() function (formerly file read and validation) has been broken out into two separate functions: read_timber() and debark().
• read_timber() and debark() are now much more permissive, and require only one field (res_outcome).
• Rather than using a 'scrap pile' for excluded resistance outcomes, sawmill now includes status fields. This functionality will be improved in future releases.
• Continued depreciation of timber versions.
• Added preliminary support for risk ratios (sawmill does not compute an odds ratio from risk ratio, but will not immediately error).

v3.0.4: improve start_mill()

This update adds support for importing .CSV files, and adds additional prompts to start_mill() to better guide the user.

v3.0.3: minor bug-fixes

Avoid errors where no identifiers are specified.

v3.0.2: Bugfix for package "here"

This release fixes a bug where sawmill would fail to install where "here" is not present on the system, as it was called without addition to namespace.

• Dropped call to here.
• Added here to suggests for future.

v3.0.1: Add local timber

This release adds local_timber -- the ability to generate timber using sawmill itself. See the README for instructions.

v3.0.0: Support v3 queries (timber; CEDAR web formtting)

The release drops support for older timber formats, and adds support for the latest version (v3; CEDAR web formatting).

It also includes some quality of life updates such as missing column handling, and an update script.

Fix pval column

The fixes the following error message in the do_MA function: "Error: Can't combine ..1$pval and ..2$pval ." By converting both ordinary pvals and pvals from meta-analysis to the character data type, we avoid mismatches in this column. R requires that all entries in a column have the same data type.

Fix format meta-analysis results error

This fixes an error with the format_ma_results function, which formats ma_results, a data frame of the parameters calculated by the meta-analysis package when developing the random-effects model. ma_results is an optional, low priority output, as the actual weighted average odds ratios needed for the models are saved in the planks, but this error was preventing the planks from being saved. The problem was that the call column must not be deleted from ma_results, otherwise the following error message occurs: "Names must be unique. These are duplicated: "b" at locations __ and __, ... beta, and se". This is not a complete fix, as the ma_results output still does not save to a csv, but this fix does allow the planks to be saved. If the scrap pile is needed, the timber will have to be run through sawmill without meta-analysis groupings.

Fix meta-analysis p-value error

Remove unnecessary as.character() conversion for the meta-analysis p-value to avoid confusing the ma_result's pvalue with the timber's pvalue. Fixes the following error message: "Error: Can't combine ..1$pval and ..2$pval ."

v2.1.1: Refresh environment

This is the seventh release, and the second version of sawmill.

This release processes timber from CEDAR v1 and v2 for use in the iAM.AMR models. It has a complete set of key features, meaning it is at feature parity with cedarr, the R package designed for CEDAR v1. However, it only performs meta-analyses for v2 timber. For a listing of these key features, please see previous release notes.

Changes from the last release include:

• The global environment is now cleared at the beginning of the processing pipeline to prevent global variables containing meta-analysis results (ma_results, ma_results_formatted) and unusable factors (scrap_pile) from carrying over to consecutive re-runs of sawmill, which may change or eliminate the contents of one or more of these global variables.