The dossier package provides tools to analyze and compare decisions made in the data analysis across multiple applied studies on the same topic. It helps explore the “garden of forking paths” by highlighting patterns, similarities, and differences in how analyses are conducted.
Installation
You can install the development version from GitHub with:
# install.packages("remotes")
remotes::install_github("huizezhang-sherry/dossier")Example
This is a basic example which shows you how to solve a common problem:
library(dossier)
library(dplyr)
library(text)
raw_df <- readr::read_csv(system.file("papers.csv", package = "dossier"))
(tbl_df <- as_decision_tbl(raw_df))
#> # A tibble: 35 × 8
#> paper variable type model method parameter reason decision
#> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <chr>
#> 1 braga barometric_pressure param… gene… LOESS smoothin… to mi… <NA>
#> 2 braga barometric_pressure spati… gene… <NA> <NA> to al… chosen …
#> 3 braga humidity param… gene… LOESS smoothin… to mi… <NA>
#> 4 braga humidity spati… gene… <NA> <NA> to al… chosen …
#> 5 braga temperature param… gene… LOESS smoothin… to mi… <NA>
#> 6 braga temperature spati… gene… <NA> <NA> to al… chosen …
#> 7 braga time param… gene… LOESS smoothin… to el… <NA>
#> 8 braga time spati… gene… <NA> <NA> seaso… chosen …
#> 9 katsouyanni humidity param… gene… LOESS smoothin… to mi… <NA>
#> 10 katsouyanni humidity tempo… gene… <NA> <NA> <NA> same da…
#> # ℹ 25 more rows
# select the variable-type pair to compare papers
count_variable_type(tbl_df)
#> # A tibble: 11 × 3
#> variable type .n
#> <chr> <chr> <int>
#> 1 temperature parameter 6
#> 2 time parameter 6
#> 3 humidity parameter 4
#> 4 temperature temporal 4
#> 5 PM temporal 3
#> 6 humidity temporal 3
#> 7 time spatial 2
#> 8 barometric_pressure parameter 1
#> 9 barometric_pressure spatial 1
#> 10 humidity spatial 1
#> 11 temperature spatial 1
(df <- tbl_df |> filter_var_type(n = 6)) # first 6 variable-type pairs
#> # A tibble: 26 × 8
#> paper variable type model method parameter reason decision
#> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <chr>
#> 1 braga humidity parameter generaliz… LOESS smoothin… to mi… <NA>
#> 2 braga temperature parameter generaliz… LOESS smoothin… to mi… <NA>
#> 3 braga time parameter generaliz… LOESS smoothin… to el… <NA>
#> 4 katsouyanni humidity parameter generaliz… LOESS smoothin… to mi… <NA>
#> 5 katsouyanni humidity temporal generaliz… <NA> <NA> <NA> same da…
#> 6 katsouyanni temperature parameter generaliz… LOESS smoothin… to mi… <NA>
#> 7 katsouyanni temperature temporal generaliz… <NA> <NA> <NA> same da…
#> 8 katsouyanni time parameter generaliz… LOESS smoothin… to mi… <NA>
#> 9 ostro PM temporal Poisson r… <NA> <NA> previ… 2-day a…
#> 10 ostro humidity parameter Poisson r… smoot… smoothin… <NA> 3
#> # ℹ 16 more rows
# subset to papers with at least 3 decisions (include all)
count_paper_decisions(df)
#> paper .n
#> 1 ostro 11
#> 2 katsouyanni 8
#> 3 schwartz 8
#> 4 zanobetti 8
#> 5 braga 6
#> 6 peel 6
(paper_df <- df |> filter_papers(n_value = 3))
#> # A tibble: 26 × 8
#> paper variable type model method parameter reason decision
#> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <chr>
#> 1 braga humidity parameter generaliz… LOESS smoothin… to mi… <NA>
#> 2 braga temperature parameter generaliz… LOESS smoothin… to mi… <NA>
#> 3 braga time parameter generaliz… LOESS smoothin… to el… <NA>
#> 4 katsouyanni humidity parameter generaliz… LOESS smoothin… to mi… <NA>
#> 5 katsouyanni humidity temporal generaliz… <NA> <NA> <NA> same da…
#> 6 katsouyanni temperature parameter generaliz… LOESS smoothin… to mi… <NA>
#> 7 katsouyanni temperature temporal generaliz… <NA> <NA> <NA> same da…
#> 8 katsouyanni time parameter generaliz… LOESS smoothin… to mi… <NA>
#> 9 ostro PM temporal Poisson r… <NA> <NA> previ… 2-day a…
#> 10 ostro humidity parameter Poisson r… smoot… smoothin… <NA> 3
#> # ℹ 16 more rows
count_paper_pair_decisions(paper_df)
#> # A data frame: 15 × 3
#> paper1 paper2 .n
#> <chr> <chr> <int>
#> 1 braga katsouyanni 6
#> 2 braga ostro 4
#> 3 braga peel 2
#> 4 braga schwartz 4
#> 5 braga zanobetti 6
#> 6 katsouyanni ostro 6
#> 7 katsouyanni peel 3
#> 8 katsouyanni schwartz 4
#> 9 katsouyanni zanobetti 8
#> 10 ostro peel 6
#> 11 ostro schwartz 7
#> 12 ostro zanobetti 6
#> 13 peel schwartz 5
#> 14 peel zanobetti 3
#> 15 schwartz zanobetti 4
# calculate the text embed
embed_df <- paper_df |> compute_text_embed()
#> Processing batch 1/1
#>
#> Completed layers output for texts (variable: 1/1, duration: 4.024554 secs).
#>
#> Completed layers aggregation for word_type_embeddings.
#>
#> Completed layers aggregation (variable 1/1, duration: 3.146090 secs).
#>
#> Completed layers aggregation (variable 1/1, duration: 2.791564 secs).
#>
#> Minutes from start: 0.174
#> Estimated embedding time left = 0 minutes
# calculate decision similarity
(distance_decision_df <- calc_decision_similarity(paper_df, embed = embed_df))
#> # A tibble: 74 × 4
#> paper1 paper2 decision dist
#> <chr> <chr> <chr> <dbl>
#> 1 braga katsouyanni humidity_parameter_method 1
#> 2 braga katsouyanni temperature_parameter_method 1
#> 3 braga katsouyanni time_parameter_method 1
#> 4 braga katsouyanni humidity_parameter_reason 0.946
#> 5 braga katsouyanni temperature_parameter_reason 0.946
#> 6 braga katsouyanni time_parameter_reason 0.760
#> 7 braga ostro humidity_parameter_method 0
#> 8 braga ostro temperature_parameter_method 0
#> 9 braga ostro time_parameter_method 0
#> 10 braga ostro time_parameter_reason 0.728
#> # ℹ 64 more rows
# aggregate from decision similarity to paper similarity
(distance_df <- distance_decision_df |> calc_paper_similarity())
#> # A tibble: 15 × 4
#> paper1 paper2 dist similarity
#> <fct> <fct> <dbl> <dbl>
#> 1 braga katsouyanni 0.0578 0.942
#> 2 braga ostro 0.818 0.182
#> 3 braga peel 1 0
#> 4 braga schwartz 0.101 0.899
#> 5 braga zanobetti 0.0546 0.945
#> 6 katsouyanni ostro 0.627 0.373
#> 7 katsouyanni peel 0.786 0.214
#> 8 katsouyanni schwartz 0.140 0.860
#> 9 katsouyanni zanobetti 0.0799 0.920
#> 10 ostro peel 0.545 0.455
#> 11 ostro schwartz 0.565 0.435
#> 12 ostro zanobetti 0.686 0.314
#> 13 peel schwartz 0.633 0.367
#> 14 peel zanobetti 0.825 0.175
#> 15 schwartz zanobetti 0.143 0.857Diagnostics and visualization
diag_decision_ppp(distance_decision_df, distance_df)
#> # A tibble: 7 × 4
#> num_decision count prop similarity
#> <int> <int> <dbl> <dbl>
#> 1 2 1 0.0667 0
#> 2 3 2 0.133 0.194
#> 3 4 4 0.267 0.699
#> 4 5 1 0.0667 0.367
#> 5 6 5 0.333 0.606
#> 6 7 1 0.0667 0.435
#> 7 8 1 0.0667 0.920
view_pairs(paper_df, distance_decision_df, "braga", "zanobetti")
#> Joining with `by = join_by(decision)`
#> # A tibble: 8 × 4
#> decision braga zanobetti dist
#> <chr> <chr> <chr> <dbl>
#> 1 humidity_parameter_method LOESS LOESS 1
#> 2 humidity_parameter_reason to minimize Akaike's Informati… to minim… 0.946
#> 3 temperature_parameter_method LOESS LOESS 1
#> 4 temperature_parameter_reason to minimize Akaike's Informati… to minim… 0.946
#> 5 time_parameter_method LOESS LOESS 1
#> 6 time_parameter_reason to eliminate seasonal patterns… to minim… 0.779
#> 7 humidity_temporal_decision <NA> previous… NA
#> 8 temperature_temporal_decision <NA> previous… NA
