bakaburg1 / baysren Goto Github PK

Once a session is fully labelled, it is possible to get possibly unrecovered papers from the set of relevant records:

• The papers cited by the relevant records: either manually with the creation of a .bib file, through packages like RefManageR which can access CrossRefs or directly through CrossRefs API (eg https://api.crossref.org/works/10.1098/rstb.1999.0425) which lists references. But not always all the required data is present (abstract may be missing and doi may be missing in the citations)
• Related papers on PubMed via RefManageR::GetPubMedRelated().
• More ideas?

The collected papers may represent a new session, to be labelled in the usual way.

A global random seed won't work: some models like BART or brms need a direct seed to be passed, and also the seed needs to change when the model ensemble is used.

The idea is to generate a new seed starting from an original one in every iteration of the ensemble.
The problem with just this approach is that the seed would be the same for each run of enrich_annotation_file().

So an even more general solution is needed.

The code was already implemented in the manuscript repo. Take it from there

Probably a formatting error

At the moment the package imports 28 packages, making it vulnerable.
As many packages as possible should be replaced by ad-hoc internal functions.

At the moment only search_pubmed and search_ieee have an argument to limit the number of downloaded records, but with different defaults (numeric() in the first and NULL in the second).
Such limit should be implemented also into search_wos too.
Similarly, also perform_search_session needs this argument to be percolated to the search functions, but it's not clear to me whether it should be applied to API search only or also when importing already downloaded records.

For some reason, the ID field keeps collecting duplicated IDs. I need to understand where those are generated and fix the problem. I suspect this happens in create_annotation_file when importing previous session data.

Probably an error in the formatting functions

The function fails if the query returns no results, with message:

Error in UseMethod("transmute") : 
no applicable method for 'transmute' applied to an object of class "NULL"

Should be easy to fix.

the function does not create a file, so it's a misnomer

Even if the function works, the following output is produced:

Unhandled promise error: argument is of length zero
Unhandled promise error: Invalid InterceptionId.. code: -32602

Understand the cause or suppress the warnings if not solvable.

res <- perform_search_session(
query = '(systematic review) AND ((heart failure) AND (COPD))',
year_query = NULL, ## <- this line triggers the error!
actions = c("API", "parsed"),
sources = c("Pubmed"),
session_name = "Session1",
query_name = "Query1",
records_folder = "Records",
overwrite = T,
skip_on_failure = FALSE,
journal = "Session_journal.csv"
)

returns:

Error in data.frame(Session_ID = session_name, Query_ID = query_name, : 
arguments imply differing number of rows: 1, 0

estimate_performance() relies on a brms model built with cmdstan.
Therefore a user would need to install both cmdstanr and cmdstan to make it work.

Either add cmdstanr and brms as Suggested packages and manage the installation, or switch to rstanarm if possible, which would also skip compilation of the stan program at every run.

Useful to interact with other bibliography tools.

Can be built manually or wrapping RefManageR or bibtex packages.

For importation, the functionality can be implemented in the import_data() function.

There duplicated indicators like "New labels"/"New_labels" and there is "Total_labeled" whose name need fixing.