Log-transform yi estimates prior to fitting MA mod #118

egouldo · 2024-08-15T03:26:51Z

update downstream processing to log-transform out-of-sample euc estimates

Originally posted by @egouldo in #88

- #102 add import tags for `log_transform()` and link to equivalent functions, apply default argument values / checks - #97 rename out argument

- equivalent to `pred_to_z()` - #102 write documentation

- #118 adapt response variable preparation to accept additional argument `dataset_log_transform` apply argument checks #116, add roxygen param #102 - #118 adapt response variable processing to accept either/or/none for dataset standardisation/log-transformation.

- #102 add import, return, and see also roxygen doc tags, replace note with details tag, rename fn doc title - #116 update argument checks conditional expression - #118 match output to `log_transform_yi()` (now returns additional cols `lower` and `upper`, not only `c("Z","VZ")`) - #118 match process to `log_transform_yi()` and #97 generalise processing to both euc/bt datasets without hard-coding dataset names in fns, and remove associated dataset-specific argument checking #116

…dardisation updates - #118 extract `lower` and `upper` transformed vals in line with addition of `log_transform_response()` / changes to `standardise_response()`

- #118 add log-transformation equivalent to `standardise_response()` and `process_resonse()` - #102 add function documentation, including examples

…pipeline - #118 call new arg `dataset_log_transform` in fn to log-transform outcomes for euc yi analysis

…ddition

- #118 ensure application of Z_VZ_preds takes the generalised colnames yi, yi_se instead of using hard-coded dataset application #97

…lied for standardising / transformation - accidentally deleted when upgrading for #118, have added creation of transform_datasets tibbles for all cases now, and then these will apply the appropriate functions in final code chunk at end

- `dat` to `data` to help with auto-matching in pmap within prepare_response_variables() wrapper #118 - ensure all family fns have ... arg for pmap application in prepare_response_variables() since all fns have different argument lengths and names

…ample code for `log_transform_response()`

… of transformation / standardisation in `standardise_response()` - #118 delete old pmap helper function

…fter #118 updates

--- For `ManyEcoEvo_yi` currently which does not have an `estimate_type` col

…ype` is missing in `ManyEcoEvo` dataframe - #118 build: devtools::document()

…ew name

---- Was returning only the first column as a vector rather than entire output of log_transform_yi() because `ifelse()` returns vectors dummy!

…ponse_variables() that now create estiamte_type col --- unnest() was throwing an error after splitting based on scenario numbers because colum estimate_type was duplicated

…erenc `VZ` colnames (i.e. from `log_transform_response()`) #118 - refactor!: `apply_VZ_exclusions()` to handle different out-of-sample colnames for each dataset #118

…s::document()`

- expose filtering of subsets to user argument: ignore outlier generation for some subsets - Expose `n_min` and `n_max` args to user - Add argument checks #116, inc. checking filter expressions for existance of col names in `data` - Conditionally slice depending on whether `n_*` args supplied in call - Define internal helper fns to reduce duplicate code - Permit multiple logical expressions in argument ignore_subsets - Allow different values of `n_min` and `n_max` for different `dataset`s bug!: properly evaluate whether arg `ignore_subsets` is null and parse to `filter_vars`, see `dplyr::dplyr_data_masking` docs!: Update docs #118 - include functional updates in roxygen docs #102 - rm targets pipeline family tag style: tidy lint #97 - linespacing / whitespace - rm TODO comments - Add section header comments

bug!: #118 ensure `estimate_type` column is added back in using argument value if existing col is replaced and stop grouping on returned data frame

egouldo · 2024-08-27T16:00:41Z

Almost there. Remaining downstream functions to fix:

meta_analyse_datasets() use same approach as n_max/n_min/VZ_cutoff argument matching strategies to deal with different outcome variables
- calculate_deviation_score()
- box_cox_transform()
- fit_sorensen_glm()
Check make_viz works as expected after updates on branch

- rm commented out `poss_fit_*()` (now exported internally in 7dabe6d) #97 - replace name of data arg to `data` #97 - switch to rlang::is_null for arg checking #116 / conditional behaviour - Add arg checking for `filter_vars`, `data`, and checking wither required columns, incl. those in `filter_vars` exist in `data` #116 - expose user-supplied arguments `outcom_variable` / `outcome_SE` and match values to arg conditions in new columns - update supply of colnames arg to meta-analysis functions docs!: update corresponding docs #102 - switch to `@importFrom` for `cli::`, add missing `@importFrom` calls - Add `@details` on user-specified exclusion of outcome variable / outcome variable SE style: spacing indentation, rm inline TODO comments #97

…118 - `fit_MA_mv()` switch from `Z_colname` to `outcome_colname` and `VZ_colname` to `outcome_SE_colname` #118 - `calculate_deviation_score()` #118: expose colname selection column to user , return meta-analytic mean in column, add `outcome_colname` to specify relevant outcome column name, add argument checking #116 - `apply_slice_conditionally()` create column `exclusion_set` if does not exist and switch to `se_log` as `VZ_colname` in example for eucalyptus` #188 - `box_cox_transform()` expose `outcome_SE_colname` arg, add relevant arg checks, update cli output, parse outcome_SE_columns` to `folded_params()` function call #188 - replace `std.error_log` arg with `se_log` for any logged `yi` data #188 - and update wrapper function calls in `meta_analyse_datasets()`, provide NULL default for new arg `outcome_SE` #188 - `fit_box_cox_ratings_cat()`: add arg checks #116, replace arg name `.data` with `data`, switch to `inject` approach as per `fit_box_cox_ratings_cont()` #97 - fit_box_cox_ratings_cont()` switch to `data` arg from `.data`, add arg checks #116 - add arg checks to `standardise_response()`, `fit_sorensen_glm()`, `fit_multivar_MA()` , `fit_uni_mixed_effects()` #116 docs!: update corresponding docs - delete obsolete `@examples` code #102 - add #TODO comments #97 - add missing `@import` calls - switch to `@importFrom` from `@import` for `purrr::` `rlang::` and `tidyr::` et al - don't import `rlang::` in package.R bug!: minor bug fixes - unquote bare colname in pointblank check for `exclude_extreme_VZ()` - `fit_metafor_mv_reduced()` replace obs_id with study_id as random effect

- add import tag for `purrr::transpose()` in `fit_multivar_MA()` #102 - defuse argument `outcome` in `cli()` call in `fit_boxcox_ratings_cont()` - `meta_analyse_datasets()`: - use correct argument in calls to `fit_boxcox_ratings_*()` (`data` from `.data`) - apply coalesce on list-cols only and `@import` `tidyselect::where()` (previously non-list-cols were ignored because character columns were explicitly grouped, however current approach in pipeline is ungrouped in order to allow for generalisability / scalability #87. Can't caoalesce a character vector and list, must be of same type). build!: update calls in targets `yi` pipeline

egouldo · 2024-08-28T14:59:29Z

huge push! make_viz() fns to check:

get_MA_fit_stats()
gg_forest()
poss_viz_funnel()

egouldo added bug an unexpected problem or unintended behavior feature a feature request or enhancement labels Aug 15, 2024

egouldo added this to the Respond Reviewer Comments milestone Aug 15, 2024

egouldo self-assigned this Aug 15, 2024

egouldo added a commit that referenced this issue Aug 15, 2024

#118 add option to log_transform() in prepare_response_variables()

d2ab529

- #102 add import tags for `log_transform()` and link to equivalent functions, apply default argument values / checks - #97 rename out argument

egouldo added a commit that referenced this issue Aug 15, 2024

#118 create log_transform_yi()

44c2e76

- equivalent to `pred_to_z()` - #102 write documentation

egouldo added a commit that referenced this issue Aug 15, 2024

#118 add function log_transform_response()

e199dfb

- #118 add log-transformation equivalent to `standardise_response()` and `process_resonse()` - #102 add function documentation, including examples

egouldo added a commit that referenced this issue Aug 15, 2024

#118 update application of prepare_response_variables() in targets …

5af7899

…pipeline - #118 call new arg `dataset_log_transform` in fn to log-transform outcomes for euc yi analysis

egouldo added a commit that referenced this issue Aug 15, 2024

#118 #97 #102 devtools::document()

000eb65

egouldo added a commit that referenced this issue Aug 15, 2024

#118 rm call to redundant arg in pred_to_Z()

8e5a7cf

egouldo added a commit that referenced this issue Aug 15, 2024

#118 delete prev. data assignment occurring in middle of unfinished a…

babbf8a

…ddition

egouldo added a commit that referenced this issue Aug 15, 2024

#118 Increment version and update description

d8ceeb6

egouldo added a commit that referenced this issue Aug 15, 2024

#118 rm redundant code erroneously left-behind

a8f5456

- #118 ensure application of Z_VZ_preds takes the generalised colnames yi, yi_se instead of using hard-coded dataset application #97

egouldo added a commit that referenced this issue Aug 15, 2024

#118 replace second .x arg with .y arg in map2() call

b22133a

egouldo added a commit that referenced this issue Aug 23, 2024

#118 assign names to fns stored in tibbles

d258f69

egouldo added a commit that referenced this issue Aug 23, 2024

#118 mv fn-specific arguments out of topic roxygen block

5ddb518

egouldo added a commit that referenced this issue Aug 23, 2024

#118 #102 add example code for standardise_response() and update ex…

a25b7ed

…ample code for `log_transform_response()`

egouldo added a commit that referenced this issue Aug 23, 2024

#118 add ... arg for using in pmap

84176be

egouldo added a commit that referenced this issue Aug 23, 2024

#118 #102 update seealso doc

75a0631

egouldo added a commit that referenced this issue Aug 23, 2024

- #118 add pmap internal helper function for differential application…

ba6cf17

… of transformation / standardisation in `standardise_response()` - #118 delete old pmap helper function

egouldo added a commit that referenced this issue Aug 23, 2024

#118 add ... arg to prepare_response_variables()

effb92b

egouldo added a commit that referenced this issue Aug 23, 2024

#118 #102 add example code

032244f

egouldo added a commit that referenced this issue Aug 23, 2024

- Update arg supply to targets call on prepare_response_variables() a…

3d45d72

…fter #118 updates

egouldo added a commit that referenced this issue Aug 23, 2024

#102 #118 devtools::document()

ee5b6fa

egouldo added a commit that referenced this issue Aug 23, 2024

#118 #102 export log_transform_response()

d409381

egouldo added a commit that referenced this issue Aug 23, 2024

#118 add check for col estimate_type existing, create it when FALSE

4a5a5fa

--- For `ManyEcoEvo_yi` currently which does not have an `estimate_type` col

egouldo added a commit that referenced this issue Aug 23, 2024

- #118 docs: Add explanation about updated behaviour when `estimate_t…

62bee48

…ype` is missing in `ManyEcoEvo` dataframe - #118 build: devtools::document()

egouldo added a commit that referenced this issue Aug 23, 2024

#118 #97 make new name for processed output and return object using n…

4ceec4d

…ew name

egouldo added a commit that referenced this issue Aug 23, 2024

#118 fix: conditional application of log_transform_yi

fe6ee93

---- Was returning only the first column as a vector rather than entire output of log_transform_yi() because `ifelse()` returns vectors dummy!

egouldo added a commit that referenced this issue Aug 24, 2024

build!: update call to apply_VZ_exclusions() after 02d058e #118

98ed2e2

egouldo added a commit that referenced this issue Aug 24, 2024

build!: increment version and devtools::document() #118 #97

9ef1127

egouldo added a commit that referenced this issue Aug 24, 2024

fix: Fix roxygen entries with missing @family tag #118 and `devtool…

5a6126c

…s::document()`

egouldo added a commit that referenced this issue Aug 27, 2024

bug!: fix cli expressions, rm improper use of glue::glue() #97

684a880

bug!: #118 ensure `estimate_type` column is added back in using argument value if existing col is replaced and stop grouping on returned data frame

egouldo added a commit that referenced this issue Aug 27, 2024

build!: update targets pipeline to reflect all #118 changes

a679cbb

egouldo removed this from the Respond Reviewer Comments milestone Aug 28, 2024

egouldo mentioned this issue Aug 28, 2024

118 log transform yi #121

Merged

egouldo linked a pull request Aug 28, 2024 that will close this issue

118 log transform yi #121

Merged

egouldo closed this as completed in #121 Aug 28, 2024

egouldo mentioned this issue Aug 28, 2024

give fit_MA_mv() a back transformation argument #66

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Log-transform yi estimates prior to fitting MA mod #118

Log-transform yi estimates prior to fitting MA mod #118

egouldo commented Aug 15, 2024 •

edited

Loading

egouldo commented Aug 27, 2024 •

edited

Loading

egouldo commented Aug 28, 2024 •

edited

Loading

Log-transform yi estimates prior to fitting MA mod #118

Log-transform yi estimates prior to fitting MA mod #118

Comments

egouldo commented Aug 15, 2024 • edited Loading

egouldo commented Aug 27, 2024 • edited Loading

egouldo commented Aug 28, 2024 • edited Loading

egouldo commented Aug 15, 2024 •

edited

Loading

egouldo commented Aug 27, 2024 •

edited

Loading

egouldo commented Aug 28, 2024 •

edited

Loading