1 指標

1AS-unitあたりの誤用数 (e_asunit)、誤用を含まないAS-unitの割合 (e_nashi_as_count_wariai)、誤用を含む節数 (e_ari_count)、誤用を含まない節数 (e_nashi_count)を使って分析した。

2 Data import

library(tidyverse)
## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr     1.1.2     ✔ readr     2.1.4
## ✔ forcats   1.0.0     ✔ stringr   1.5.0
## ✔ ggplot2   3.4.2     ✔ tibble    3.2.1
## ✔ lubridate 1.9.2     ✔ tidyr     1.3.0
## ✔ purrr     1.0.1     
## ── 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 errors
library(ggpubr)
library(rstatix)
## 
## Attaching package: 'rstatix'
## 
## The following object is masked from 'package:stats':
## 
##     filter
library(readxl)

dat <- read_excel("/Users/riku/Library/CloudStorage/Dropbox/zemizemi/paper/caf/data_accuracy/accuracy0525.xlsx", sheet = "Sheet1")


dat$id <- as.factor(dat$id)
dat$trueid <- as.factor(dat$trueid)

dat$task <- ordered(dat$task, levels=c("L","M","H"))
dat$proficiency <- as.factor(dat$proficiency)
dat$proficiency <- ordered(dat$proficiency, levels=c("lower","middle","upper"))

3 Boxplot

3.0.1 Box plot ———–

bxp <- ggboxplot(
  dat, x = "task", y = "e_asunit", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_asunit")

bxp <- ggboxplot(
  dat, x = "task", y = "e_nashi_as_count_wariai", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_nashi_as_count_wariai")

bxp <- ggboxplot(
  dat, x = "task", y = "e_ari_count", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_ari_count")

bxp <- ggboxplot(
  dat, x = "task", y = "e_nashi_count", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_nashi_count")

bxp <- ggboxplot(
  dat, x = "task", y = "e_nashi_count_wariai", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_nashi_count_wariai")

4 Comparison between M and H ——-

dat_MH <- dat %>% filter(task == "M"| task == "H")
dat_MH$task <- ordered(dat_MH$task, levels=c("M","H"))

4.1 e_asunit MH ——-

4.1.1 Box plot ———–

bxp <- ggboxplot(
  dat_MH, x = "task", y = "e_asunit", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_asunit")

4.1.2 Check assumptions ——-

dat_MH %>%
  group_by(task, proficiency) %>%
  identify_outliers(e_asunit)
## # A tibble: 3 × 11
##   task  proficiency id    trueid e_asunit e_nashi_as_count_wariai e_ari_count
##   <ord> <ord>       <fct> <fct>     <dbl>                   <dbl>       <dbl>
## 1 M     upper       9     11         2.25                    0.25           4
## 2 H     middle      2     3          2.11                    0             11
## 3 H     middle      24    24         2.5                     0              6
## # ℹ 4 more variables: e_nashi_count <dbl>, e_nashi_count_wariai <dbl>,
## #   is.outlier <lgl>, is.extreme <lgl>
dat_MH %>%
  group_by(task, proficiency) %>%
  shapiro_test(e_asunit)
## # A tibble: 6 × 5
##   task  proficiency variable statistic       p
##   <ord> <ord>       <chr>        <dbl>   <dbl>
## 1 M     lower       e_asunit     0.899 0.181  
## 2 M     middle      e_asunit     0.871 0.0678 
## 3 M     upper       e_asunit     0.922 0.302  
## 4 H     lower       e_asunit     0.927 0.378  
## 5 H     middle      e_asunit     0.779 0.00542
## 6 H     upper       e_asunit     0.932 0.402
ggqqplot(dat_MH, "e_asunit", ggtheme = theme_bw()) +
  facet_grid(task ~ proficiency)

dat_MH %>%
  group_by(task) %>%
  levene_test(e_asunit ~ proficiency)
## # A tibble: 2 × 5
##   task    df1   df2 statistic     p
##   <ord> <int> <int>     <dbl> <dbl>
## 1 M         2    32     0.831 0.445
## 2 H         2    32     0.863 0.431

###remove trueid 3 24 for extreme outlier

dat_MH_easunit <- dat_MH %>% filter(!(trueid %in% c("3", "24")))

dat_MH_easunit %>%
  group_by(task, proficiency) %>%
  identify_outliers(e_asunit)
## # A tibble: 1 × 11
##   task  proficiency id    trueid e_asunit e_nashi_as_count_wariai e_ari_count
##   <ord> <ord>       <fct> <fct>     <dbl>                   <dbl>       <dbl>
## 1 M     upper       9     11         2.25                    0.25           4
## # ℹ 4 more variables: e_nashi_count <dbl>, e_nashi_count_wariai <dbl>,
## #   is.outlier <lgl>, is.extreme <lgl>
dat_MH_easunit %>%
  group_by(task, proficiency) %>%
  shapiro_test(e_asunit)
## # A tibble: 6 × 5
##   task  proficiency variable statistic      p
##   <ord> <ord>       <chr>        <dbl>  <dbl>
## 1 M     lower       e_asunit     0.899 0.181 
## 2 M     middle      e_asunit     0.848 0.0543
## 3 M     upper       e_asunit     0.922 0.302 
## 4 H     lower       e_asunit     0.927 0.378 
## 5 H     middle      e_asunit     0.935 0.500 
## 6 H     upper       e_asunit     0.932 0.402
ggqqplot(dat_MH_easunit, "e_asunit", ggtheme = theme_bw()) +
  facet_grid(task ~ proficiency)

dat_MH_easunit %>%
  group_by(task) %>%
  levene_test(e_asunit ~ proficiency)
## # A tibble: 2 × 5
##   task    df1   df2 statistic     p
##   <ord> <int> <int>     <dbl> <dbl>
## 1 M         2    30     0.592 0.560
## 2 H         2    30     1.64  0.212

4.1.3 Computation ——-

res.aov <- anova_test(
  data = dat_MH_easunit, dv = e_asunit, wid = trueid,
  between = proficiency, within = task
)
get_anova_table(res.aov)
## ANOVA Table (type III tests)
## 
##             Effect DFn DFd      F        p p<.05   ges
## 1      proficiency   2  30  0.320 0.728000       0.016
## 2             task   1  30 14.135 0.000736     * 0.107
## 3 proficiency:task   2  30  0.670 0.519000       0.011
interaction.plot(x.factor = dat_MH_easunit$task, trace.factor = dat_MH_easunit$proficiency, 
                 response = dat_MH_easunit$e_asunit, fun = mean,
                 type = "b", legend = TRUE, trace.label = "TASK")

### main effect

dat_MH_easunit %>%
  pairwise_t_test(
    e_asunit ~ task, paired = TRUE, 
    p.adjust.method = "bonferroni"
  )
## # A tibble: 1 × 10
##   .y.     group1 group2    n1    n2 statistic    df       p   p.adj p.adj.signif
## * <chr>   <chr>  <chr>  <int> <int>     <dbl> <dbl>   <dbl>   <dbl> <chr>       
## 1 e_asun… M      H         33    33      3.79    32 6.37e-4 6.37e-4 ***

M > H

4.2 e_nashi_as_count_wariai MH ——-

4.2.1 Box plot ———–

bxp <- ggboxplot(
  dat_MH, x = "task", y = "e_nashi_as_count_wariai", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_nashi_as_count_wariai")

4.2.2 Check assumptions ——-

dat_MH %>%
  group_by(task, proficiency) %>%
  identify_outliers(e_nashi_as_count_wariai)
## # A tibble: 2 × 11
##   task  proficiency id    trueid e_asunit e_nashi_as_count_wariai e_ari_count
##   <ord> <ord>       <fct> <fct>     <dbl>                   <dbl>       <dbl>
## 1 M     middle      16    15            0                       1           0
## 2 M     upper       4     6             0                       1           0
## # ℹ 4 more variables: e_nashi_count <dbl>, e_nashi_count_wariai <dbl>,
## #   is.outlier <lgl>, is.extreme <lgl>
dat_MH %>%
  group_by(task, proficiency) %>%
  shapiro_test(e_nashi_as_count_wariai)
## # A tibble: 6 × 5
##   task  proficiency variable                statistic     p
##   <ord> <ord>       <chr>                       <dbl> <dbl>
## 1 M     lower       e_nashi_as_count_wariai     0.885 0.119
## 2 M     middle      e_nashi_as_count_wariai     0.934 0.423
## 3 M     upper       e_nashi_as_count_wariai     0.946 0.584
## 4 H     lower       e_nashi_as_count_wariai     0.918 0.303
## 5 H     middle      e_nashi_as_count_wariai     0.940 0.498
## 6 H     upper       e_nashi_as_count_wariai     0.933 0.418
ggqqplot(dat_MH, "e_nashi_as_count_wariai", ggtheme = theme_bw()) +
  facet_grid(task ~ proficiency)

dat_MH %>%
  group_by(task) %>%
  levene_test(e_nashi_as_count_wariai ~ proficiency)
## # A tibble: 2 × 5
##   task    df1   df2 statistic     p
##   <ord> <int> <int>     <dbl> <dbl>
## 1 M         2    32     0.153 0.859
## 2 H         2    32     0.777 0.468

4.2.3 Computation ——-

res.aov <- anova_test(
  data = dat_MH, dv = e_nashi_as_count_wariai, wid = trueid,
  between = proficiency, within = task
)
get_anova_table(res.aov)
## ANOVA Table (type III tests)
## 
##             Effect DFn DFd     F     p p<.05   ges
## 1      proficiency   2  32 0.224 0.801       0.009
## 2             task   1  32 2.842 0.102       0.027
## 3 proficiency:task   2  32 1.183 0.320       0.023
interaction.plot(x.factor = dat_MH$task, trace.factor = dat_MH$proficiency, 
                 response = dat_MH$e_nashi_as_count_wariai, fun = mean,
                 type = "b", legend = TRUE, trace.label = "TASK")

4.3 e_ari_count MH ——-

4.3.1 Box plot ———–

bxp <- ggboxplot(
  dat_MH, x = "task", y = "e_ari_count", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_ari_count")

4.3.2 Check assumptions ——-

dat_MH %>%
  group_by(task, proficiency) %>%
  identify_outliers(e_ari_count)
## # A tibble: 3 × 11
##   task  proficiency id    trueid e_asunit e_nashi_as_count_wariai e_ari_count
##   <ord> <ord>       <fct> <fct>     <dbl>                   <dbl>       <dbl>
## 1 M     upper       4     6          0                      1               0
## 2 H     lower       21    20         1                      0.364          10
## 3 H     middle      2     3          2.11                   0              11
## # ℹ 4 more variables: e_nashi_count <dbl>, e_nashi_count_wariai <dbl>,
## #   is.outlier <lgl>, is.extreme <lgl>
dat_MH %>%
  group_by(task, proficiency) %>%
  shapiro_test(e_ari_count)
## # A tibble: 6 × 5
##   task  proficiency variable    statistic      p
##   <ord> <ord>       <chr>           <dbl>  <dbl>
## 1 M     lower       e_ari_count     0.865 0.0674
## 2 M     middle      e_ari_count     0.956 0.728 
## 3 M     upper       e_ari_count     0.861 0.0505
## 4 H     lower       e_ari_count     0.828 0.0218
## 5 H     middle      e_ari_count     0.860 0.0485
## 6 H     upper       e_ari_count     0.910 0.212
ggqqplot(dat_MH, "e_ari_count", ggtheme = theme_bw()) +
  facet_grid(task ~ proficiency)

dat_MH %>%
  group_by(task) %>%
  levene_test(e_ari_count ~ proficiency)
## # A tibble: 2 × 5
##   task    df1   df2 statistic     p
##   <ord> <int> <int>     <dbl> <dbl>
## 1 M         2    32     1.98  0.154
## 2 H         2    32     0.124 0.884

4.3.3 Computation ——-

res.aov <- anova_test(
  data = dat_MH, dv = e_ari_count, wid = trueid,
  between = proficiency, within = task
)
get_anova_table(res.aov)
## ANOVA Table (type III tests)
## 
##             Effect DFn DFd     F     p p<.05   ges
## 1      proficiency   2  32 0.100 0.905       0.004
## 2             task   1  32 1.158 0.290       0.010
## 3 proficiency:task   2  32 0.062 0.940       0.001
interaction.plot(x.factor = dat_MH$task, trace.factor = dat_MH$proficiency, 
                 response = dat_MH$e_ari_count, fun = mean,
                 type = "b", legend = TRUE, trace.label = "TASK")

4.4 e_nashi_count MH ——-

4.4.1 Box plot ———–

bxp <- ggboxplot(
  dat_MH, x = "task", y = "e_nashi_count", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_nashi_count")

4.4.2 Check assumptions ——-

dat_MH %>%
  group_by(task, proficiency) %>%
  identify_outliers(e_nashi_count)
## # A tibble: 4 × 11
##   task  proficiency id    trueid e_asunit e_nashi_as_count_wariai e_ari_count
##   <ord> <ord>       <fct> <fct>     <dbl>                   <dbl>       <dbl>
## 1 M     middle      3     2         0.333                   0.667           3
## 2 M     middle      11    9         0.2                     0.8             1
## 3 M     middle      10    12        1                       0.4             4
## 4 M     middle      15    16        1.25                    0.25            4
## # ℹ 4 more variables: e_nashi_count <dbl>, e_nashi_count_wariai <dbl>,
## #   is.outlier <lgl>, is.extreme <lgl>
dat_MH %>%
  group_by(task, proficiency) %>%
  shapiro_test(e_nashi_count)
## # A tibble: 6 × 5
##   task  proficiency variable      statistic      p
##   <ord> <ord>       <chr>             <dbl>  <dbl>
## 1 M     lower       e_nashi_count     0.868 0.0722
## 2 M     middle      e_nashi_count     0.886 0.106 
## 3 M     upper       e_nashi_count     0.851 0.0378
## 4 H     lower       e_nashi_count     0.907 0.223 
## 5 H     middle      e_nashi_count     0.927 0.351 
## 6 H     upper       e_nashi_count     0.929 0.366
ggqqplot(dat_MH, "e_nashi_count", ggtheme = theme_bw()) +
  facet_grid(task ~ proficiency)

dat_MH %>%
  group_by(task) %>%
  levene_test(e_nashi_count ~ proficiency)
## # A tibble: 2 × 5
##   task    df1   df2 statistic      p
##   <ord> <int> <int>     <dbl>  <dbl>
## 1 M         2    32     4.39  0.0207
## 2 H         2    32     0.832 0.445

4.4.3 Computation ——-

res.aov <- anova_test(
  data = dat_MH, dv = e_nashi_count, wid = trueid,
  between = proficiency, within = task
)
get_anova_table(res.aov)
## ANOVA Table (type III tests)
## 
##             Effect DFn DFd     F     p p<.05   ges
## 1      proficiency   2  32 3.838 0.032     * 0.150
## 2             task   1  32 3.062 0.090       0.024
## 3 proficiency:task   2  32 1.136 0.334       0.018
interaction.plot(x.factor = dat_MH$task, trace.factor = dat_MH$proficiency, 
                 response = dat_MH$e_nashi_count, fun = mean,
                 type = "b", legend = TRUE, trace.label = "TASK")

dat_MH %>%
  pairwise_t_test(
    e_nashi_count ~ proficiency, 
    p.adjust.method = "bonferroni"
  )
## # A tibble: 3 × 9
##   .y.           group1 group2    n1    n2       p p.signif   p.adj p.adj.signif
## * <chr>         <chr>  <chr>  <int> <int>   <dbl> <chr>      <dbl> <chr>       
## 1 e_nashi_count lower  middle    22    24 0.494   ns       1       ns          
## 2 e_nashi_count lower  upper     22    24 0.00225 **       0.00676 **          
## 3 e_nashi_count middle upper     24    24 0.0132  *        0.0397  *

Upper > Lower

Upper > Middle

5 Comparison between L and M ——-

dat_LM <- dat %>% filter(task == "L"| task == "M")
dat_LM$task <- ordered(dat_LM$task, levels=c("L","M"))

5.1 e_asunit LM ——-

5.1.1 Box plot ———–

bxp <- ggboxplot(
  dat_LM, x = "task", y = "e_asunit", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_asunit")

5.1.2 Check assumptions ——-

dat_LM %>%
  group_by(task, proficiency) %>%
  identify_outliers(e_asunit)
## # A tibble: 2 × 11
##   task  proficiency id    trueid e_asunit e_nashi_as_count_wariai e_ari_count
##   <ord> <ord>       <fct> <fct>     <dbl>                   <dbl>       <dbl>
## 1 L     middle      24    24         3.25                    0              8
## 2 M     upper       9     11         2.25                    0.25           4
## # ℹ 4 more variables: e_nashi_count <dbl>, e_nashi_count_wariai <dbl>,
## #   is.outlier <lgl>, is.extreme <lgl>
dat_LM %>%
  group_by(task, proficiency) %>%
  shapiro_test(e_asunit)
## # A tibble: 6 × 5
##   task  proficiency variable statistic      p
##   <ord> <ord>       <chr>        <dbl>  <dbl>
## 1 L     lower       e_asunit     0.972 0.906 
## 2 L     middle      e_asunit     0.816 0.0143
## 3 L     upper       e_asunit     0.897 0.144 
## 4 M     lower       e_asunit     0.899 0.181 
## 5 M     middle      e_asunit     0.871 0.0678
## 6 M     upper       e_asunit     0.922 0.302
ggqqplot(dat_LM, "e_asunit", ggtheme = theme_bw()) +
  facet_grid(task ~ proficiency)

dat_LM %>%
  group_by(task) %>%
  levene_test(e_asunit ~ proficiency)
## # A tibble: 2 × 5
##   task    df1   df2 statistic     p
##   <ord> <int> <int>     <dbl> <dbl>
## 1 L         2    32     0.942 0.400
## 2 M         2    32     0.831 0.445
# REMOVE outlier
dat_LM_easunit <- dat_LM %>% filter(!(trueid %in% c("11","24")))

5.1.3 Computation ——-

res.aov <- anova_test(
  data = dat_LM_easunit, dv = e_asunit, wid = trueid,
  between = proficiency, within = task
)
get_anova_table(res.aov)
## ANOVA Table (type II tests)
## 
##             Effect DFn DFd     F     p p<.05      ges
## 1      proficiency   2  30 0.336 0.717       0.017000
## 2             task   1  30 0.045 0.833       0.000328
## 3 proficiency:task   2  30 1.324 0.281       0.019000
interaction.plot(x.factor = dat_LM_easunit$task, trace.factor = dat_LM_easunit$proficiency, 
                 response = dat_LM_easunit$e_asunit, fun = mean,
                 type = "b", legend = TRUE, trace.label = "TASK")

5.2 e_nashi_as_count_wariai LM ——-

5.2.1 Box plot ———–

bxp <- ggboxplot(
  dat_LM, x = "task", y = "e_nashi_as_count_wariai", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_nashi_as_count_wariai")

5.2.2 Check assumptions ——-

dat_LM %>%
  group_by(task, proficiency) %>%
  identify_outliers(e_nashi_as_count_wariai)
## # A tibble: 2 × 11
##   task  proficiency id    trueid e_asunit e_nashi_as_count_wariai e_ari_count
##   <ord> <ord>       <fct> <fct>     <dbl>                   <dbl>       <dbl>
## 1 M     middle      16    15            0                       1           0
## 2 M     upper       4     6             0                       1           0
## # ℹ 4 more variables: e_nashi_count <dbl>, e_nashi_count_wariai <dbl>,
## #   is.outlier <lgl>, is.extreme <lgl>
dat_LM %>%
  group_by(task, proficiency) %>%
  shapiro_test(e_nashi_as_count_wariai)
## # A tibble: 6 × 5
##   task  proficiency variable                statistic     p
##   <ord> <ord>       <chr>                       <dbl> <dbl>
## 1 L     lower       e_nashi_as_count_wariai     0.922 0.335
## 2 L     middle      e_nashi_as_count_wariai     0.986 0.998
## 3 L     upper       e_nashi_as_count_wariai     0.928 0.362
## 4 M     lower       e_nashi_as_count_wariai     0.885 0.119
## 5 M     middle      e_nashi_as_count_wariai     0.934 0.423
## 6 M     upper       e_nashi_as_count_wariai     0.946 0.584
ggqqplot(dat_LM, "e_nashi_as_count_wariai", ggtheme = theme_bw()) +
  facet_grid(task ~ proficiency)

dat_LM %>%
  group_by(task) %>%
  levene_test(e_nashi_as_count_wariai ~ proficiency)
## # A tibble: 2 × 5
##   task    df1   df2 statistic     p
##   <ord> <int> <int>     <dbl> <dbl>
## 1 L         2    32     1.22  0.308
## 2 M         2    32     0.153 0.859

5.2.3 Computation ——-

res.aov <- anova_test(
  data = dat_LM, dv = e_nashi_as_count_wariai, wid = trueid,
  between = proficiency, within = task
)
get_anova_table(res.aov)
## ANOVA Table (type III tests)
## 
##             Effect DFn DFd     F     p p<.05   ges
## 1      proficiency   2  32 0.206 0.815       0.009
## 2             task   1  32 3.372 0.076       0.028
## 3 proficiency:task   2  32 0.111 0.895       0.002
interaction.plot(x.factor = dat_LM$task, trace.factor = dat_LM$proficiency, 
                 response = dat_LM$e_nashi_as_count_wariai, fun = mean,
                 type = "b", legend = TRUE, trace.label = "TASK")

TASKは有意傾向

5.3 e_ari_count LM ——-

5.3.1 Box plot ———–

bxp <- ggboxplot(
  dat_LM, x = "task", y = "e_ari_count", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_ari_count")

5.3.2 Check assumptions ——-

dat_LM %>%
  group_by(task, proficiency) %>%
  identify_outliers(e_ari_count)
## # A tibble: 4 × 11
##   task  proficiency id    trueid e_asunit e_nashi_as_count_wariai e_ari_count
##   <ord> <ord>       <fct> <fct>     <dbl>                   <dbl>       <dbl>
## 1 L     lower       14    13        0.286                   0.714           2
## 2 L     lower       30    31        1                       0.2             8
## 3 L     lower       32    33        0.333                   0.667           1
## 4 M     upper       4     6         0                       1               0
## # ℹ 4 more variables: e_nashi_count <dbl>, e_nashi_count_wariai <dbl>,
## #   is.outlier <lgl>, is.extreme <lgl>
dat_LM %>%
  group_by(task, proficiency) %>%
  shapiro_test(e_ari_count)
## # A tibble: 6 × 5
##   task  proficiency variable    statistic      p
##   <ord> <ord>       <chr>           <dbl>  <dbl>
## 1 L     lower       e_ari_count     0.923 0.340 
## 2 L     middle      e_ari_count     0.923 0.308 
## 3 L     upper       e_ari_count     0.955 0.708 
## 4 M     lower       e_ari_count     0.865 0.0674
## 5 M     middle      e_ari_count     0.956 0.728 
## 6 M     upper       e_ari_count     0.861 0.0505
ggqqplot(dat_LM, "e_ari_count", ggtheme = theme_bw()) +
  facet_grid(task ~ proficiency)

dat_LM %>%
  group_by(task) %>%
  levene_test(e_ari_count ~ proficiency)
## # A tibble: 2 × 5
##   task    df1   df2 statistic      p
##   <ord> <int> <int>     <dbl>  <dbl>
## 1 L         2    32      2.90 0.0695
## 2 M         2    32      1.98 0.154

5.3.3 Computation ——-

res.aov <- anova_test(
  data = dat_LM, dv = e_ari_count, wid = trueid,
  between = proficiency, within = task
)
get_anova_table(res.aov)
## ANOVA Table (type III tests)
## 
##             Effect DFn DFd     F     p p<.05   ges
## 1      proficiency   2  32 0.098 0.907       0.004
## 2             task   1  32 1.130 0.296       0.011
## 3 proficiency:task   2  32 0.302 0.742       0.006
interaction.plot(x.factor = dat_LM$task, trace.factor = dat_LM$proficiency, 
                 response = dat_LM$e_ari_count, fun = mean,
                 type = "b", legend = TRUE, trace.label = "TASK")

5.4 e_nashi_count LM ——-

5.4.1 Box plot ———–

bxp <- ggboxplot(
  dat_LM, x = "task", y = "e_nashi_count", add = "jitter",
  color = "proficiency", palette = "lancet"
)
bxp + ggtitle("e_nashi_count")

5.4.2 Check assumptions ——-

dat_LM %>%
  group_by(task, proficiency) %>%
  identify_outliers(e_nashi_count)
## # A tibble: 5 × 11
##   task  proficiency id    trueid e_asunit e_nashi_as_count_wariai e_ari_count
##   <ord> <ord>       <fct> <fct>     <dbl>                   <dbl>       <dbl>
## 1 L     upper       1     1         0.333                   0.667           3
## 2 M     middle      3     2         0.333                   0.667           3
## 3 M     middle      11    9         0.2                     0.8             1
## 4 M     middle      10    12        1                       0.4             4
## 5 M     middle      15    16        1.25                    0.25            4
## # ℹ 4 more variables: e_nashi_count <dbl>, e_nashi_count_wariai <dbl>,
## #   is.outlier <lgl>, is.extreme <lgl>
dat_LM %>%
  group_by(task, proficiency) %>%
  shapiro_test(e_nashi_count)
## # A tibble: 6 × 5
##   task  proficiency variable      statistic      p
##   <ord> <ord>       <chr>             <dbl>  <dbl>
## 1 L     lower       e_nashi_count     0.916 0.289 
## 2 L     middle      e_nashi_count     0.942 0.529 
## 3 L     upper       e_nashi_count     0.864 0.0553
## 4 M     lower       e_nashi_count     0.868 0.0722
## 5 M     middle      e_nashi_count     0.886 0.106 
## 6 M     upper       e_nashi_count     0.851 0.0378
ggqqplot(dat_LM, "e_nashi_count", ggtheme = theme_bw()) +
  facet_grid(task ~ proficiency)

dat_LM %>%
  group_by(task) %>%
  levene_test(e_nashi_count ~ proficiency)
## # A tibble: 2 × 5
##   task    df1   df2 statistic      p
##   <ord> <int> <int>     <dbl>  <dbl>
## 1 L         2    32      1.20 0.313 
## 2 M         2    32      4.39 0.0207

5.4.3 Computation ——-

res.aov <- anova_test(
  data = dat_LM, dv = e_nashi_count, wid = trueid,
  between = proficiency, within = task
)
get_anova_table(res.aov)
## ANOVA Table (type III tests)
## 
##             Effect DFn DFd     F     p p<.05      ges
## 1      proficiency   2  32 3.089 0.059       1.21e-01
## 2             task   1  32 0.003 0.953       3.13e-05
## 3 proficiency:task   2  32 1.729 0.194       3.00e-02
interaction.plot(x.factor = dat_LM$task, trace.factor = dat_LM$proficiency, 
                 response = dat_LM$e_nashi_count, fun = mean,
                 type = "b", legend = TRUE, trace.label = "TASK")

dat_LM %>%
  pairwise_t_test(
    e_nashi_count ~ proficiency, 
    p.adjust.method = "bonferroni"
  )
## # A tibble: 3 × 9
##   .y.           group1 group2    n1    n2       p p.signif  p.adj p.adj.signif
## * <chr>         <chr>  <chr>  <int> <int>   <dbl> <chr>     <dbl> <chr>       
## 1 e_nashi_count lower  middle    22    24 0.309   ns       0.926  ns          
## 2 e_nashi_count lower  upper     22    24 0.00464 **       0.0139 *           
## 3 e_nashi_count middle upper     24    24 0.0558  ns       0.167  ns

6 まとめ

6.1 1AS-unitあたりの誤用数 (e_asunit)

MH M > H

LM 認知負荷も習熟度も有意ではなかった。

6.2 誤用を含まないAS-unitの割合 (e_nashi_as_count_wariai)

MH 認知負荷も習熟度も有意ではなかった。

LM タスクは有意傾向(p = .076), L < M

6.3 誤用を含まない節数 (e_nashi_count)

MHでは、習熟度が有意、Lower < Upper; Middle < Upper,

LMでは、習熟度が有意傾向(p = .059), Lower < Upper。

6.4 まとめのまとめ

6.4.1 1AS-unitあたりの誤用数

内在性負荷が上がると、習熟度に関わらず、1AS-unitあたりの誤用数が少なくなる。

外在性負荷ではそういう変化がない。

内在性負荷が高いタスクがより複雑であるため、参加者がそのタスクに集中し、その結果として誤用数が減少した可能性がある。 外在性負荷と誤用数との間に直接的な関連性はないかも？

6.4.2 誤用を含まないAS-unitの割合

外在性・内在性負荷を問わず、習熟度に関わらず、誤用を含まないAS-unitの割合は負荷の高い方が高かくなる。 高負荷タスクでは参加者がより注意深くなり、その結果誤用を含まないAS-unitが増えた。

？？シンプルな文を言ってしまう？？他の指標との照合が必要。

6.4.3 誤用を含まない節数

MHでは、Middle < Upperですが、LMではなかった。

中位群 (middle) は、外在性負荷が上がっても誤用を含まない節数は減らないが、内在性負荷が上がったら、上位群 (upper) に比べて誤用を含まない節数が少なくなる。

習熟度が高い参加者が内在性負荷の高いタスクに対してより適応しているかも。