Using rcldf to analyse datasets:
To show you how to use rcldf to analyse datasets, we’re
going to test whether languages that distinguish inclusive pronouns from
exclusive pronouns (i.e. [clusivity][https://en.wikipedia.org/wiki/Clusivity]), also tend to
have high rigidity in social structure.
To do this, we will use the Grambank Feature GB028: Is there a distinction between inclusive and exclusive?, and the D-PLACE Trait EA113: Degree of rigidity in social structures.
First, let’s get the published version of Grambank off Zenodo. It’s always a good idea to use the published version as this is a citable and versioned product which makes replicating your analysis easier for other researchers. To do this, go to the Zenodo page for Grambank, and find the download link under the ‘Files’ section and copy it.
It will look like this https://zenodo.org/records/7844558/files/grambank/grambank-v1.0.3.zip?download=1.
Give that to rcldf:
grambank <- cldf("https://zenodo.org/records/7844558/files/grambank/grambank-v1.0.3.zip?download=1")
grambank
#> A CLDF dataset with 6 tables (CodeTable, contributors.csv, families.csv, LanguageTable, ParameterTable, ValueTable)
#>
#> Skirgård, Hedvig and Haynie, Hannah J. and Blasi, Damián E. and Hammarström, Harald and Collins, Jeremy and Latarche, Jay J. and Lesage, Jakob and Weber, Tobias and Witzlack-Makarevich, Alena and Passmore, Sam and Chira, Angela and Maurits, Luke and Dinnage, Russell and Dunn, Michael and Reesink, Ger and Singer, Ruth and Bowern, Claire and Epps, Patience and Hill, Jane and Vesakoski, Outi and Robbeets, Martine and Abbas, Noor Karolin and Auer, Daniel and Bakker, Nancy A. and Barbos, Giulia and Borges, Robert D. and Danielsen, Swintha and Dorenbusch, Luise and Dorn, Ella and Elliott, John and Falcone, Giada and Fischer, Jana and Ghanggo Ate, Yustinus and Gibson, Hannah and Göbel, Hans-Philipp and Goodall, Jemima A. and Gruner, Victoria and Harvey, Andrew and Hayes, Rebekah and Heer, Leonard and Herrera Miranda, Roberto E. and Hübler, Nataliia and Huntington-Rainey, Biu and Ivani, Jessica K. and Johns, Marilen and Just, Erika and Kashima, Eri and Kipf, Carolina and Klingenberg, Janina V. and König, Nikita and Koti, Aikaterina and Kowalik, Richard G. A. and Krasnoukhova, Olga and Lindvall, Nora L.M. and Lorenzen, Mandy and Lutzenberger, Hannah and Martins, Tônia R.A. and Mata German, Celia and van der Meer, Suzanne and Montoya Samamé, Jaime and Müller, Michael and Muradoglu, Saliha and Neely, Kelsey and Nickel, Johanna and Norvik, Miina and Oluoch, Cheryl Akinyi and Peacock, Jesse and Pearey, India O.C. and Peck, Naomi and Petit, Stephanie and Pieper, Sören and Poblete, Mariana and Prestipino, Daniel and Raabe, Linda and Raja, Amna and Reimringer, Janis and Rey, Sydney C. and Rizaew, Julia and Ruppert, Eloisa and Salmon, Kim K. and Sammet, Jill and Schembri, Rhiannon and Schlabbach, Lars and Schmidt, Frederick W.P. and Skilton, Amalia and Smith, Wikaliler Daniel and de Sousa, Hilário and Sverredal, Kristin and Valle, Daniel and Vera, Javier and Voß, Judith and Witte, Tim and Wu, Henry and Yam, Stephanie and Ye 葉婧婷, Jingting and Yong, Maisie and Yuditha, Tessa and Zariquiey, Roberto and Forkel, Robert and Evans, Nicholas and Levinson, Stephen C. and Haspelmath, Martin and Greenhill, Simon J. and Atkinson, Quentin D. and Gray, Russell D. (in press). Grambank reveals the importance of genealogical constraints on linguistic diversity and highlights the impact of language loss. Science Advances.Now, let’s use dplyr to get the variable we want from
grambank. We use summary to see what the dataset looks
like. It is a CLDF ‘Structure’ Dataset with six tables:
summary(grambank)
#> A Cross-Linguistic Data Format (CLDF) dataset:
#> Name: Grambank v1.0
#> Path: /Users/simon/Library/Caches/org.R-project.R/R/rcldf/zenodo_records_7844558_files_grambank_grambank_v1_0_3_zip_5ba67f1e8557c79b5e4fffaab8f58bcc/grambank-grambank-7ae000c/cldf
#> Type: http://cldf.clld.org/v1.0/terms.rdf#StructureDataset
#> Tables:
#> 1/6: CodeTable (4 columns, 398 rows)
#> 2/6: contributors.csv (5 columns, 139 rows)
#> 3/6: families.csv (2 columns, 215 rows)
#> 4/6: LanguageTable (13 columns, 2467 rows)
#> 5/6: ParameterTable (12 columns, 195 rows)
#> 6/6: ValueTable (9 columns, 441663 rows)
#> Sources: 0
#> Cite:
#> Skirgård, Hedvig and Haynie, Hannah J. and Blasi, Damián E. and Hammarström, Harald and Collins, Jeremy and Latarche, Jay J. and Lesage, Jakob and Weber, Tobias and Witzlack-Makarevich, Alena and Passmore, Sam and Chira, Angela and Maurits, Luke and Dinnage, Russell and Dunn, Michael and Reesink, Ger and Singer, Ruth and Bowern, Claire and Epps, Patience and Hill, Jane and Vesakoski, Outi and Robbeets, Martine and Abbas, Noor Karolin and Auer, Daniel and Bakker, Nancy A. and Barbos, Giulia and Borges, Robert D. and Danielsen, Swintha and Dorenbusch, Luise and Dorn, Ella and Elliott, John and Falcone, Giada and Fischer, Jana and Ghanggo Ate, Yustinus and Gibson, Hannah and Göbel, Hans-Philipp and Goodall, Jemima A. and Gruner, Victoria and Harvey, Andrew and Hayes, Rebekah and Heer, Leonard and Herrera Miranda, Roberto E. and Hübler, Nataliia and Huntington-Rainey, Biu and Ivani, Jessica K. and Johns, Marilen and Just, Erika and Kashima, Eri and Kipf, Carolina and Klingenberg, Janina V. and König, Nikita and Koti, Aikaterina and Kowalik, Richard G. A. and Krasnoukhova, Olga and Lindvall, Nora L.M. and Lorenzen, Mandy and Lutzenberger, Hannah and Martins, Tônia R.A. and Mata German, Celia and van der Meer, Suzanne and Montoya Samamé, Jaime and Müller, Michael and Muradoglu, Saliha and Neely, Kelsey and Nickel, Johanna and Norvik, Miina and Oluoch, Cheryl Akinyi and Peacock, Jesse and Pearey, India O.C. and Peck, Naomi and Petit, Stephanie and Pieper, Sören and Poblete, Mariana and Prestipino, Daniel and Raabe, Linda and Raja, Amna and Reimringer, Janis and Rey, Sydney C. and Rizaew, Julia and Ruppert, Eloisa and Salmon, Kim K. and Sammet, Jill and Schembri, Rhiannon and Schlabbach, Lars and Schmidt, Frederick W.P. and Skilton, Amalia and Smith, Wikaliler Daniel and de Sousa, Hilário and Sverredal, Kristin and Valle, Daniel and Vera, Javier and Voß, Judith and Witte, Tim and Wu, Henry and Yam, Stephanie and Ye 葉婧婷, Jingting and Yong, Maisie and Yuditha, Tessa and Zariquiey, Roberto and Forkel, Robert and Evans, Nicholas and Levinson, Stephen C. and Haspelmath, Martin and Greenhill, Simon J. and Atkinson, Quentin D. and Gray, Russell D. (in press). Grambank reveals the importance of genealogical constraints on linguistic diversity and highlights the impact of language loss. Science Advances.Let’s start by extracting a list of languages in this dataset:
languages <- grambank$tables$LanguageTable |>
select(ID, Name, Macroarea, Latitude, Longitude)
# only selecting some columns above to make it easier to see
languages
#> # A tibble: 2,467 × 5
#> ID Name Macroarea Latitude Longitude
#> <chr> <chr> <chr> <dbl> <dbl>
#> 1 abad1241 Abadi Papunesia -9.03 147.
#> 2 abar1238 Mungbam Africa 6.58 10.2
#> 3 abau1245 Abau Papunesia -3.97 141.
#> 4 abee1242 Abé Africa 5.60 -4.38
#> 5 aben1249 Abenlen Ayta Papunesia 15.4 120.
#> 6 abip1241 Abipon South America -29 -61
#> 7 abkh1244 Abkhaz Eurasia 43.1 41.2
#> 8 abua1245 Abu' Arapesh Papunesia -3.46 143.
#> 9 abui1241 Abui Papunesia -8.31 125.
#> 10 abun1252 Abun Papunesia -0.571 132.
#> # ℹ 2,457 more rowsWe can also look at the ParameterTable to see
information on our parameter of interest: “GB028: Is there a distinction
between inclusive and exclusive?”. The ID of this feature
is “GB028”, so let’s just see that one:
grambank$tables$ParameterTable |> filter(ID=='GB028')
#> # A tibble: 1 × 12
#> ID Name Description ColumnSpec Patrons Grambank_ID_desc Boundness
#> <chr> <chr> <chr> <chr> <chr> <chr> <chr>
#> 1 GB028 Is there a di… "## Is the… <NA> HJH GB028 Clusivity <NA>
#> # ℹ 5 more variables: Flexivity <chr>, Gender_or_Noun_Class <chr>,
#> # Locus_of_Marking <chr>, Word_Order <chr>, Informativity <chr>Ok, now let’s get all the Values for this Parameter in
ValueTable. Following the cldf standard means
that the Parameter will be mapped into the ValueTable in the
Parameter_ID column, so let’s select all the rows in
ValueTable that match Parameter_ID of GB028:
values <- grambank$tables$ValueTable |>
filter(Parameter_ID=='GB028') |>
select(ID, Language_ID, Parameter_ID, Value, Source)Now we need to get the data from D-PLACE. We’ll use the Github repository to get these data to show you how that works, but you should probably use the published version on Zenodo for proper work. Get the github repository link https://github.com/D-PLACE/dplace-data and give that to rcldf too:
dplace <- cldf("https://github.com/D-PLACE/dplace-cldf")
summary(dplace)
#> A Cross-Linguistic Data Format (CLDF) dataset:
#> Name: D-PLACE aggregated dataset
#> Path: /Users/simon/Library/Caches/org.R-project.R/R/rcldf/github_D_PLACE_dplace_cldf_266fef649c0d27ffcdc6502c39bad1fa/D-PLACE-dplace-cldf-c00e58b/cldf
#> Type: http://cldf.clld.org/v1.0/terms.rdf#StructureDataset
#> Tables:
#> 1/7: CodeTable (5 columns, 15450 rows)
#> 2/7: ContributionTable (8 columns, 122 rows)
#> 3/7: LanguageTable (20 columns, 6174 rows)
#> 4/7: MediaTable (6 columns, 109 rows)
#> 5/7: ParameterTable (11 columns, 2987 rows)
#> 6/7: TreeTable (9 columns, 109 rows)
#> 7/7: ValueTable (11 columns, 631668 rows)
#> Sources: 0
#> Cite:
#> Kathryn R. Kirby, Russell D. Gray, Simon J. Greenhill, Fiona M. Jordan, Stephanie Gomes-Ng, Hans-Jörg Bibiko, Damián E. Blasi, Carlos A. Botero, Claire Bowern, Carol R. Ember, Dan Leehr, Bobbi S. Low, Joe McCarter, William Divale, and Michael C. Gavin. (2016). D-PLACE: A Global Database of Cultural, Linguistic and Environmental Diversity. PLoS ONE, 11(7): e0158391. doi:10.1371/journal.pone.0158391.Great, let’s get the values for the parameter we want. D-PLACE
indexes the variables in a column Var_ID:
dplace$tables$ValueTable |> filter(Var_ID=='EA113')
#> # A tibble: 1,291 × 11
#> ID Soc_ID Var_ID Value Code_ID Comment Source sub_case year
#> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <chr>
#> 1 ea-117482 Aa1 EA113 Flexible EA113-2 <NA> <NA> Nyai Nyae reg… 1950
#> 2 ea-117483 Aa2 EA113 <NA> EA113-NA <NA> <NA> <NA> <NA>
#> 3 ea-117484 Aa3 EA113 <NA> EA113-NA <NA> <NA> <NA> <NA>
#> 4 ea-117485 Aa4 EA113 <NA> EA113-NA <NA> <NA> <NA> <NA>
#> 5 ea-117486 Aa5 EA113 Flexible EA113-2 <NA> <NA> Epulu net-hun… 1930
#> 6 ea-117487 Aa6 EA113 <NA> EA113-NA <NA> <NA> <NA> <NA>
#> 7 ea-117488 Aa7 EA113 <NA> EA113-NA <NA> <NA> <NA> <NA>
#> 8 ea-117489 Aa8 EA113 <NA> EA113-NA <NA> <NA> <NA> <NA>
#> 9 ea-117490 Aa9 EA113 <NA> EA113-NA <NA> <NA> <NA> <NA>
#> 10 ea-117491 Ab1 EA113 <NA> EA113-NA <NA> <NA> <NA> <NA>
#> # ℹ 1,281 more rows
#> # ℹ 2 more variables: source_coded_data <chr>, admin_comment <chr>We want to merge this with our Grambank data but need a way to do
this. D-PLACE stores glottocode information that maps each
society to a language. This column is also in the Languages table in
Grambank. So, let’s get the trait values for EA113 and add the
glottocodes from LanguageTable using a join:
# get languages from DPLACE
dplanguages <- dplace$tables$LanguageTable |> select(ID, Glottocode)
# get values for EA113 and merge with language information
ea113 <- dplace$tables$ValueTable |>
filter(Var_ID=='EA113') |>
select(Soc_ID, Value) |>
left_join(dplanguages, join_by(Soc_ID==ID))
# rename `Value` to EA113
ea113 <- ea113 |> mutate(EA113=Value) |> select(Glottocode, EA113)
head(ea113)
#> # A tibble: 6 × 2
#> Glottocode EA113
#> <chr> <chr>
#> 1 juho1239 Flexible
#> 2 okie1245 <NA>
#> 3 nama1265 <NA>
#> 4 dama1270 <NA>
#> 5 bila1255 Flexible
#> 6 sand1273 <NA>Now get the grambank data into the same format:
gb028 <- values |>
mutate(Glottocode=Language_ID, GB028=Value) |>
select(Glottocode, GB028)
head(gb028)
#> # A tibble: 6 × 2
#> Glottocode GB028
#> <chr> <chr>
#> 1 abad1241 1
#> 2 abar1238 0
#> 3 abau1245 0
#> 4 abee1242 0
#> 5 aben1249 1
#> 6 abip1241 0…and finally join them up using the mutual column ‘Glottocode’. I’ll
use an inner join here to only get the languages/societies that are in
both datasets. And we’ll use na.omit to only keep rows that
have data for both variables:
df <- gb028 |> inner_join(ea113) |> na.omit()
#> Joining with `by = join_by(Glottocode)`
head(df)
#> # A tibble: 6 × 3
#> Glottocode GB028 EA113
#> <chr> <chr> <chr>
#> 1 bila1255 0 Flexible
#> 2 chig1238 0 Flexible
#> 3 fefe1239 0 Flexible
#> 4 gand1255 1 Rigid
#> 5 hehe1240 0 Flexible
#> 6 juku1254 0 RigidOk, it looks like we only have 17 rows for this pairing. That’s a little small, but let’s plot the data:
p1 <- ggplot(df, aes(x=GB028)) + geom_histogram(stat='count')
#> Warning in geom_histogram(stat = "count"): Ignoring unknown parameters:
#> `binwidth`, `bins`, and `pad`
p2 <- ggplot(df, aes(x=EA113)) + geom_histogram(stat='count')
#> Warning in geom_histogram(stat = "count"): Ignoring unknown parameters:
#> `binwidth`, `bins`, and `pad`
p1 / p2 # patchwork
Hmm. It looks like societies that do not mark this distinction between inclusive and exclusive tend to be those with flexible social structure. Looking good for our hypothesis, but let’s test it formally to make sure we’re not seeing a chance pattern:
tab <- table(df$GB028, df$EA113)
tab
#>
#> Flexible Rigid
#> 0 9 4
#> 1 3 1
chisq.test(tab)
#> Warning in chisq.test(tab): Chi-squared approximation may be incorrect
#>
#> Pearson's Chi-squared test with Yates' continuity correction
#>
#> data: tab
#> X-squared = 4.986e-31, df = 1, p-value = 1….and the probability here is 1.0 of getting a result like this due to chance. So yes, this is just a chance result and there is no evidence in these data that languages which distinguish inclusive from exclusive also tend to be those that have rigid social structures. There goes our big paper in Nature/Science.