German MDA
Andressa Costa
April 14, 2024
FACTOR ANALYSIS
with frequencies normed by 100 and minimum text length of 100 words
Features used in the factor analysis: 1. adja: attributive adjective, 2. adjd: predicative adjective, 3. adv: adverb, 4. artdef: definite article, 5. artdem: demonstrative article, 6. artindef: indefinite article, 7. artinter: interrogative article, 8. artposs: possessive article, 9. attindef: attributive indefinita, 10. awl: average word length, 11. dass: conjunction dass/that, 12. diskmark: discourse marker, 13. eignam: proper name, 14. gen:gentive, 15. gespptk: conversational particle, 16. kaus: causal clause, 17. kokom: comparative clause, 18. kokonj: coordinating conjunction, 19. kommv: verb of communication, 20. kond: conditional clause, 21. nabs: abstract noun, 22. nebfin: subordinating with finite verb, 23. nebzuinf: subordinating with zu infinitive, 24. noml: nominalization, 25. nreg: regular noun, 26. pass: passive, 27. pdem: demonstrative pronoun, 28. pindef: indefinite pronoun, 29. pinter: interrogative pronoun, 30. pper1: 1st person pronoun, 31. pper2: 2nd person pronoun, 32. pper3: 3rd person pronoun, 33. pposs: possessive pronoun, 34. prefl: reflexive pronoun, 35. prel: relative pronoun, 36. prep: preposition, 37. proadv: pronominal adverb, 38. ptkifg: intensity particle, 39. ptkma: modal particle, 40. ptkneg: negation particle, 41. qutag: question tag, 42. temp: temporal clause, 43. ttr: Type-token-ratio, 44. v2s: verb 2nd subordinate clause, 45. vfut: future, 46. vimp: imperative, 47. vinfzu: Zu Infinitive object clause, 48. vipast: indicative past, 49. vipres: indicative present, 50. vmod: modal verb, 51. vperf: Perfekt, 52. vpqp: Past perfect simple, 53. vspast: subjunctive past, 54. vspres: subjjunctive present, 55. vvipres: full verb indicative present
library(tidyverse)## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr 1.1.4 ✔ readr 2.1.5
## ✔ forcats 1.0.0 ✔ stringr 1.5.1
## ✔ ggplot2 3.4.4 ✔ tibble 3.2.1
## ✔ lubridate 1.9.3 ✔ tidyr 1.3.1
## ✔ purrr 1.0.2
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errorslibrary(performance)
library(psych)##
## Attaching package: 'psych'
##
## The following objects are masked from 'package:ggplot2':
##
## %+%, alphalibrary(ggcorrplot)# Load the data
features_count_normed_100 <- read_csv("features_extraction/features_count_normed_100.csv")## Rows: 3068 Columns: 56
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): doc_id
## dbl (55): awl, ttr, pindef, pposs, prefl, pdem, pinter, prel, artdef, artind...
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.Put the column ‘register’ after ‘doc_id’
features_count_normed_100 <- features_count_normed_100 %>%
select(doc_id, awl:eignam)Get the indices for the correlation matrix
indices_100 <- features_count_normed_100 %>%
select(-doc_id)Add the column register to tha data
features_count_normed_100$register <- gsub("\\d+", "", features_count_normed_100$doc_id)
features_count_normed_100 <- features_count_normed_100 %>%
mutate(register = recode(register, ag = "everyday conversation", agw = "academic article humanities",
bk = "business communication", bl = "blog", bt = "fiction literature",
bz = "package insert", df = "documentary", dgw = "dissertation humanities",
dinw = "dissertation ing. nat. sci.", dlw = "dissertation life sciences",
essay = "essay", ev = "expert lecture", fd = "movie drama", fs = "tv series",
ga = "instruction manual", gn = "spoken news", gt = "legal texts",
hk = "horoscope", ik = "institutional communication", int = "interview",
jl = "young literature", km = "comedy", ks = "crime series", la = "editorial",
lb = "letter to the editor", ld = "song lyrics", lk = "readers comment",
ma = "magazine article", md = "news", mn = "opinion", pc = "professional chat",
pd = "sermon", pg = "oral exam", pinw = "popular science articles ing nat sci",
plw = "popular science articles life sci", pp = "plenary minutes",
pr = "political speech", rz = "recipe", sa = "job offer",
sb = "non-fiction book", sv = "student lecture", th = "theatre play",
tt = "ted talk", vl = "lectures", wa = "wikipedia article",
wd = "wikipedia talk", wlbgw = "scientific textbook humanities",
wlbinw = "scientif textbook Ing. nat. sci",
wlblw = "scientific textbook life sci", ws = "website",
za = "newspaper article", zi = "newspaper interview"))1. Evaluate the data
- Correlation Matrix
cor_matrix <- cor(indices_100) %>%
data.frame() %>%
rownames_to_column()
# Remove overcorrelated features > 0.7
write.csv(cor(indices_100)>0.7, file="factor_analysis/suspect_correlations.csv")
overcorrelated <- c("pdem", "dass", "vpqp", "gen", "vvipres", "ptkma")
indices_100 <- indices_100[ ,!(names(indices_100) %in% overcorrelated)]
write.csv(indices_100, "factor_analysis/indices_100.csv")
# Visualize the correlation matrix
cor_matrix <- round(cor(indices_100), 2)- Bartellt’s Test of Sphericity
r <- cor(indices_100)
check_sphericity_bartlett(r, n = nrow(indices_100))## # Test of Sphericity
##
## Bartlett's test of sphericity suggests that there is sufficient significant correlation in the data for factor analysis (Chisq(1176) = 71811.22, p < .001).cortest.bartlett(indices_100)## R was not square, finding R from data## $chisq
## [1] 71811.22
##
## $p.value
## [1] 0
##
## $df
## [1] 1176- KMO
check_kmo(r, n = nrow(indices_100))## # KMO Measure of Sampling Adequacy
##
## The Kaiser, Meyer, Olkin (KMO) overall measure of sampling adequacy suggests that data seems appropriate for factor analysis (KMO = 0.87). The individual KMO scores are: awl (0.72), ttr (0.66), pindef (0.96), pposs (0.72), prefl (0.81), pinter (0.96), prel (0.86), artdef (0.90), artindef (0.80), attindef (0.85), artposs (0.77), artdem (0.85), artinter (0.70), vmod (0.92), kommv (0.93), prep (0.92), kokonj (0.64), vinfzu (0.79), nebfin (0.80), kond (0.80), kaus (0.69), temp (0.51), nebzuinf (0.72), kokom (0.81), vipres (0.80), vipast (0.40*), vspres (0.65), vspast (0.90), vimp (0.85), vfut (0.81), pass (0.78), vperf (0.86), proadv (0.74), adv (0.95), pper1 (0.92), pper2 (0.87), pper3 (0.50), adja (0.96), adjd (0.93), v2s (0.74), nabs (0.81), noml (0.92), nreg (0.93), ptkneg (0.97), gespptk (0.89), ptkifg (0.94), diskmark (0.86), qutag (0.91), eignam (0.72).KMO(indices_100)## Kaiser-Meyer-Olkin factor adequacy
## Call: KMO(r = indices_100)
## Overall MSA = 0.87
## MSA for each item =
## awl ttr pindef pposs prefl pinter prel artdef
## 0.72 0.66 0.96 0.72 0.81 0.96 0.86 0.90
## artindef attindef artposs artdem artinter vmod kommv prep
## 0.80 0.85 0.77 0.85 0.70 0.92 0.93 0.92
## kokonj vinfzu nebfin kond kaus temp nebzuinf kokom
## 0.64 0.79 0.80 0.80 0.69 0.51 0.72 0.81
## vipres vipast vspres vspast vimp vfut pass vperf
## 0.80 0.40 0.65 0.90 0.85 0.81 0.78 0.86
## proadv adv pper1 pper2 pper3 adja adjd v2s
## 0.74 0.95 0.92 0.87 0.50 0.96 0.93 0.74
## nabs noml nreg ptkneg gespptk ptkifg diskmark qutag
## 0.81 0.92 0.93 0.97 0.89 0.94 0.86 0.91
## eignam
## 0.722. Determine number of factors
Scree plot
scree(indices_100, pc = FALSE)
Parallel Analysis (Parallel analysis suggests that the number of factors = 12)
fa.parallel(indices_100, fa="fa")
## Parallel analysis suggests that the number of factors = 12 and the number of components = NAGet eigenvalues
ev <- eigen(cor(indices_100))
ev$values## [1] 10.8390622 4.0320761 2.6974669 2.1463225 1.8329937 1.4916142
## [7] 1.4688990 1.3047009 1.2260808 1.1480259 1.1136819 1.0504540
## [13] 1.0161183 0.9543365 0.9325872 0.8700500 0.8358260 0.8195085
## [19] 0.7667178 0.7412144 0.7046914 0.6864539 0.6633438 0.6608061
## [25] 0.6303353 0.5939525 0.5757835 0.5665558 0.5276989 0.4930476
## [31] 0.4773794 0.4534827 0.4407766 0.4217900 0.4006770 0.3738711
## [37] 0.3567733 0.3527625 0.3228193 0.2932031 0.2699016 0.2634757
## [43] 0.2376521 0.2209981 0.1962426 0.1700112 0.1542033 0.1300776
## [49] 0.07349713. Extract and rotate the factors
Solution with 5 factors
fa_5 <- fa(indices_100, fm = "pa", nfactors = 5, rotate = "promax")## Loading required namespace: GPArotationprint(fa_5, cut = 0)## Factor Analysis using method = pa
## Call: fa(r = indices_100, nfactors = 5, rotate = "promax", fm = "pa")
## Standardized loadings (pattern matrix) based upon correlation matrix
## PA1 PA4 PA2 PA5 PA3 h2 u2 com
## awl -0.33 -0.09 -0.13 0.15 -0.04 0.227 0.77 2.0
## ttr 0.10 -0.05 -0.01 0.30 0.05 0.083 0.92 1.3
## pindef 0.60 0.18 0.12 -0.09 0.00 0.559 0.44 1.3
## pposs 0.07 0.01 -0.02 -0.06 -0.01 0.014 0.99 2.2
## prefl 0.08 0.00 0.58 -0.16 0.07 0.354 0.65 1.2
## pinter 0.25 0.43 -0.12 -0.12 0.10 0.469 0.53 2.1
## prel -0.10 -0.06 0.60 0.01 0.13 0.415 0.59 1.2
## artdef -0.32 -0.16 0.14 0.49 0.22 0.587 0.41 2.7
## artindef -0.05 -0.13 0.25 0.24 0.15 0.199 0.80 3.2
## attindef 0.06 0.17 0.31 -0.01 0.02 0.153 0.85 1.6
## artposs -0.23 0.27 0.32 -0.52 -0.01 0.385 0.62 2.7
## artdem 0.14 0.01 0.39 0.29 -0.03 0.267 0.73 2.2
## artinter 0.07 0.17 0.17 0.16 -0.13 0.104 0.90 4.2
## vmod 0.27 0.52 0.11 0.14 0.06 0.487 0.51 1.8
## kommv 0.41 0.22 0.00 -0.09 0.36 0.579 0.42 2.6
## prep -0.60 -0.18 0.13 0.23 0.12 0.674 0.33 1.7
## kokonj -0.09 -0.17 0.04 -0.33 -0.57 0.389 0.61 1.9
## vinfzu -0.11 0.14 0.65 -0.02 -0.07 0.448 0.55 1.2
## nebfin 0.52 0.17 0.43 0.03 -0.09 0.554 0.45 2.3
## kond 0.20 0.49 0.14 0.00 -0.18 0.390 0.61 1.8
## kaus 0.48 -0.05 0.10 0.12 -0.12 0.201 0.80 1.4
## temp 0.06 -0.25 0.29 -0.36 0.07 0.197 0.80 2.9
## nebzuinf -0.09 0.02 0.41 -0.06 -0.06 0.165 0.83 1.2
## kokom 0.08 0.08 0.24 -0.09 0.06 0.102 0.90 2.0
## vipres 0.23 0.79 -0.06 0.07 -0.17 0.808 0.19 1.3
## vipast 0.16 -0.55 0.32 -0.31 0.44 0.551 0.45 3.4
## vspres 0.00 -0.05 -0.01 0.03 0.43 0.178 0.82 1.0
## vspast 0.42 0.08 0.20 0.04 0.21 0.337 0.66 2.1
## vimp 0.02 0.41 -0.31 -0.26 0.13 0.442 0.56 2.9
## vfut -0.08 0.29 0.07 -0.01 0.02 0.080 0.92 1.3
## pass -0.10 0.06 0.01 0.35 -0.01 0.146 0.85 1.2
## vperf 0.16 0.30 0.04 0.14 0.19 0.213 0.79 2.8
## proadv 0.08 -0.04 0.33 0.07 -0.13 0.129 0.87 1.5
## adv 0.75 0.09 -0.08 -0.01 0.05 0.674 0.33 1.1
## pper1 0.30 0.38 -0.09 -0.36 0.09 0.681 0.32 3.1
## pper2 0.05 0.50 -0.23 -0.28 0.06 0.511 0.49 2.1
## pper3 -0.09 0.27 0.09 -0.11 -0.20 0.100 0.90 2.7
## adja -0.45 -0.22 0.16 0.29 -0.09 0.630 0.37 2.7
## adjd 0.74 -0.08 0.00 -0.15 -0.23 0.564 0.44 1.3
## v2s -0.09 0.10 -0.03 -0.17 0.45 0.275 0.73 1.5
## nabs -0.30 0.12 0.34 0.26 0.01 0.332 0.67 3.2
## noml -0.45 0.20 0.13 0.22 0.20 0.288 0.71 2.5
## nreg -0.61 -0.24 -0.15 -0.01 -0.25 0.726 0.27 1.8
## ptkneg 0.30 0.47 0.10 -0.07 0.10 0.548 0.45 2.0
## gespptk 0.76 0.01 -0.41 0.21 0.04 0.683 0.32 1.7
## ptkifg 0.78 0.00 0.20 0.08 -0.02 0.601 0.40 1.1
## diskmark 0.85 -0.20 -0.09 0.29 -0.13 0.552 0.45 1.4
## qutag 0.45 0.03 -0.30 0.11 0.09 0.294 0.71 2.0
## eignam -0.30 -0.10 -0.08 0.02 0.55 0.353 0.65 1.7
##
## PA1 PA4 PA2 PA5 PA3
## SS loadings 7.26 4.05 3.13 2.30 1.95
## Proportion Var 0.15 0.08 0.06 0.05 0.04
## Cumulative Var 0.15 0.23 0.29 0.34 0.38
## Proportion Explained 0.39 0.22 0.17 0.12 0.10
## Cumulative Proportion 0.39 0.61 0.77 0.90 1.00
##
## With factor correlations of
## PA1 PA4 PA2 PA5 PA3
## PA1 1.00 0.52 -0.04 -0.32 0.17
## PA4 0.52 1.00 0.09 -0.26 0.15
## PA2 -0.04 0.09 1.00 0.21 0.15
## PA5 -0.32 -0.26 0.21 1.00 -0.23
## PA3 0.17 0.15 0.15 -0.23 1.00
##
## Mean item complexity = 2
## Test of the hypothesis that 5 factors are sufficient.
##
## df null model = 1176 with the objective function = 23.55 with Chi Square = 71811.22
## df of the model are 941 and the objective function was 6.34
##
## The root mean square of the residuals (RMSR) is 0.04
## The df corrected root mean square of the residuals is 0.05
##
## The harmonic n.obs is 3068 with the empirical chi square 12757.92 with prob < 0
## The total n.obs was 3068 with Likelihood Chi Square = 19318.09 with prob < 0
##
## Tucker Lewis Index of factoring reliability = 0.674
## RMSEA index = 0.08 and the 90 % confidence intervals are 0.079 0.081
## BIC = 11763
## Fit based upon off diagonal values = 0.97
## Measures of factor score adequacy
## PA1 PA4 PA2 PA5 PA3
## Correlation of (regression) scores with factors 0.97 0.95 0.92 0.90 0.89
## Multiple R square of scores with factors 0.94 0.91 0.85 0.81 0.79
## Minimum correlation of possible factor scores 0.88 0.82 0.70 0.62 0.58Factorial composition for a 5 solution
FACTOR 1
positive pole: diskmark 0.85, ptkifg 0.78, gespptk 0.76, adv 0.75, adjd 0.74, pindef 0.60, nebfin 0.52, kaus 0.48, qutag 0.45, vspast 0.42, kommv 0.41, ptkneg 0.30, pper1 0.30
negative pole: nreg -0.61, prep -0.60, adja -0.45, noml -0.45, awl -0.33, artdef -0.32, nabs -0.30, eignam -0.30
FACTOR 2
positive pole: vipres 0.79, vmod 0.52, pper2 0.50, kond 0.49, ptkneg 0.47, pinter 0.43, vimp 0.41, pper1 0.38, vperf 0.30
negative pole: vipast -0.55
FACTOR 3
positive pole: vinfzu 0.65, prel 0.60, prefl 0.58, nebfin 0.43, nebzuinf 0.41, artdem 0.39, nabs 0.34, proadv 0.33, artposs 0.32, vipast 0.32, attindef 0.31
negative pole: vimp -0.31, qutag -0.30
FACTOR 4
positive pole: artdef 0.49, pass 0.35, ttr 0.30
negative pole: artposs -0.52, temp -0.36, pper1 -0.36, kokonj -0.33, vipast -0.31
FACTOR 5
positive pole: eignam 0.55, v2s 0.45, vipast 0.44, vspres 0.43, kommv 0.36
4. Compute z-scores
z-scores for each text
df_std_100 <- features_count_normed_100 %>%
select(awl:eignam)
z_scores_100 <- sapply(df_std_100, function(df_std_100) (df_std_100 - mean(df_std_100))/sd(df_std_100))
z_scores_df_100 <- as.data.frame(z_scores_100)
z_scores_df_100<- z_scores_df_100 %>%
cbind(doc_id=features_count_normed_100$doc_id,
register=features_count_normed_100$register)z-scores per register
Extract the prefixs of doc_id and save them in the new column register_id
z_scores_df_100$register_id <- gsub("\\d+", "", z_scores_df_100$doc_id)
z_scores_reg_100 <- z_scores_df_100 %>%
group_by(register_id) %>%
select(awl:eignam) %>%
summarise_all(sum)## Adding missing grouping variables: `register_id`register <- c('everyday conversation', 'academic article humanities', 'business communication',
'blog', 'fiction literature', 'package insert', 'documentary', 'dissertation humanities',
'dissertation ing. nat. sci.', 'dissertation life sciences', 'essay', 'expert lecture',
'movie drama', 'tv series', 'instruction manual', 'spoken news', 'legal texts',
'horoscope', 'institutional communication', 'interview', 'young literature', 'comedy',
'crime series', 'editorial', 'letter to the editor', 'song lyrics', 'readers comment',
'magazine article', 'news', 'opinion', 'professional chat', 'sermon', 'oral exam',
'popular science articles ing. nat. sci.', 'popular science articles life sci',
'plenary minutes', 'political speech', 'recipe', 'job offer', 'non-fiction book',
'student lecture', 'theatre play', 'ted talk', 'lectures', 'wikipedia article',
'wikipedia talk', 'scientific textbook humanities', 'scientif textbook Ing. nat. sci',
'scientific textbook life sci', 'website', 'newspaper article', 'newspaper interview')
z_scores_reg_100$register <- register
# Reorder the columns
z_scores_reg_100 <- z_scores_reg_100 %>%
select(register_id, register, everything())5. Compute factor scores
Select the variables of each factor with z-scores
f1_100 <- z_scores_df_100[c("doc_id", "register", "awl", "prep", "adja", "noml", "nreg",
"pindef", "kommv", "kaus", "vspast", "adv", "adjd", "gespptk",
"ptkifg", "diskmark", "qutag", "nebfin")]
f2_100 <- z_scores_df_100[c("doc_id", "register", "pinter", "vmod", "nebfin", "kond", "vipres",
"vimp", "vperf", "pper1", "pper2", "ptkneg", "vipast")]
f3_100 <- z_scores_df_100[c("doc_id", "register", "prefl", "prel", "attindef", "artdem", "vinfzu",
"nebzuinf", "proadv", "nabs")]
f4_100 <- z_scores_df_100[c("doc_id", "register", "ttr", "artdef", "pass", "artposs", "temp", "kokonj")]
f5_100 <- z_scores_df_100[c("doc_id", "register", "vspres", "v2s", "eignam")]Add the positive loadinsg-features and subtract the negative-loading features on each dimension. (Features that loaded in many dimensions where only used to compute the scores for the dimension with higher loading)
df_scores_100 <- z_scores_df_100 %>%
group_by(doc_id) %>%
mutate(f1_score_100 = sum(c(pindef, kommv, kaus, vspast, adv, adjd, gespptk, ptkifg, diskmark, qutag,
nebfin)) - sum(c(awl, prep, adja, noml, nreg)),
f2_score_100 = sum(c(pinter, vmod, nebfin, kond, vipres, vimp, vperf, pper1, pper2, ptkneg)) -
vipast,
f3_score_100 = sum(c(prefl, prel, attindef, artdem, vinfzu, nebzuinf, proadv, nabs)),
f4_score_100 = sum(c(ttr, artdef, pass)) - sum(c(artposs, temp, kokonj)),
f5_score_100 = sum(c(vspres, v2s, eignam))
)
df_scores_100 %>%
select(register_id:f5_score_100)## Adding missing grouping variables: `doc_id`## # A tibble: 3,068 × 7
## # Groups: doc_id [3,068]
## doc_id register_id f1_score_100 f2_score_100 f3_score_100 f4_score_100
## <chr> <chr> <dbl> <dbl> <dbl> <dbl>
## 1 ag001 ag 20.0 5.65 -6.04 1.07
## 2 ag002 ag 28.3 10.9 -7.57 0.154
## 3 ag003 ag 18.4 8.12 -8.06 5.56
## 4 ag004 ag 19.3 6.62 -7.48 1.29
## 5 ag005 ag 25.6 10.5 -6.84 -0.0117
## 6 ag006 ag 24.2 9.15 -7.18 2.87
## 7 ag007 ag 26.4 11.0 -4.23 0.701
## 8 ag008 ag 17.9 7.62 -5.41 0.544
## 9 ag009 ag 20.0 9.96 -3.10 -0.247
## 10 ag010 ag 32.7 8.01 -5.56 -0.00374
## # ℹ 3,058 more rows
## # ℹ 1 more variable: f5_score_100 <dbl>Count total number of texts per registers
text_count_reg_100 <- df_scores_100 %>%
group_by(register_id) %>%
summarize(count = n()) %>%
rename(text_count = count)
# Add column register from df_scores to text_count_register
text_count_reg_100$register <- registerReorder columns
text_count_reg_100 <- text_count_reg_100 %>%
select(register_id, register, text_count)
print(text_count_reg_100)## # A tibble: 52 × 3
## register_id register text_count
## <chr> <chr> <int>
## 1 ag everyday conversation 50
## 2 agw academic article humanities 50
## 3 bk business communication 50
## 4 bl blog 100
## 5 bt fiction literature 75
## 6 bz package insert 50
## 7 df documentary 50
## 8 dgw dissertation humanities 50
## 9 dinw dissertation ing. nat. sci. 50
## 10 dlw dissertation life sciences 50
## # ℹ 42 more rows6. Mean dimensions scores
Computer mean dimension scores = factor_score / total texts of register Compute total factor scores for each register
df_scores2 <- df_scores_100 %>%
group_by(register_id) %>%
summarise(f1_score_total = sum(f1_score_100, na.rm=TRUE),
f2_score_total = sum(f2_score_100, na.rm=TRUE),
f3_score_total = sum(f3_score_100, na.rm=TRUE),
f4_score_total = sum(f4_score_100, na.rm=TRUE),
f5_score_total = sum(f5_score_100, na.rm=TRUE))
mean_ds_100 <- data_frame()## Warning: `data_frame()` was deprecated in tibble 1.1.0.
## ℹ Please use `tibble()` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.mean_ds_100 <- full_join(text_count_reg_100, df_scores2, by = "register_id")Compute mean dimension scores
mean_ds_100 <- mean_ds_100 %>%
group_by(register_id) %>%
mutate(mean_ds1 = f1_score_total/text_count,
mean_ds2 = f2_score_total/text_count,
mean_ds3 = f3_score_total/text_count,
mean_ds4 = f4_score_total/text_count,
mean_ds5 = f5_score_total/text_count)
mean_ds_100 %>%
select(register_id, register, mean_ds1:mean_ds5)## # A tibble: 52 × 7
## # Groups: register_id [52]
## register_id register mean_ds1 mean_ds2 mean_ds3 mean_ds4 mean_ds5
## <chr> <chr> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 ag everyday conversati… 26.4 7.77 -5.69 0.517 -0.943
## 2 agw academic article hu… -6.45 -4.29 3.34 3.03 -0.806
## 3 bk business communicat… -10.4 -3.98 0.908 0.666 -1.22
## 4 bl blog 1.21 2.01 3.62 -1.43 0.0985
## 5 bt fiction literature 3.85 -2.93 0.629 -2.75 1.12
## 6 bz package insert -8.04 3.35 -2.59 0.605 -1.04
## 7 df documentary 8.49 6.03 -0.586 -0.648 0.181
## 8 dgw dissertation humani… -6.01 -4.06 3.10 8.96 -0.588
## 9 dinw dissertation ing. n… -8.02 -4.92 1.81 9.39 -1.45
## 10 dlw dissertation life s… -9.62 -7.12 -0.942 6.19 -1.17
## # ℹ 42 more rows