Bilingual Idiom Final Analysis
library(dplyr)
library(ggplot2)
library(Hmisc)
library(lme4)
library(afex)
library(knitr)
library(data.table)
library(zoo) #replace missing with col mean
library(lmerTest)Data Preprocessing
df<-read.csv('revision_3.csv')
#convert numerical cols to factors
cols<-c('participant','prime','end','gender','idiom_lang','nat_LAN','phrase_condition','list')
df[cols]<-lapply(df[cols],factor)
colnames(df)[colnames(df)=="idiom_lang"] <- "Idiom_Language"
colnames(df)[colnames(df)=="phrase_condition"] <- "Phrase_Type"
levels(df$Phrase_Type)[levels(df$Phrase_Type)=="Experimental"] <- 'Idiomatic'
levels(df$nat_LAN)[levels(df$nat_LAN)=="2"] <- 'Mandarin-English Bilinguals'
levels(df$nat_LAN)[levels(df$nat_LAN)=="1"] <- 'Native English Monolinguals'
#replacing missing numerical values with col mean
cols1<-c('EN_rating','CN_rating','freq_HAL','age')
df[cols1] <- sapply(na.aggregate(df[cols1]),as.numeric)
head(df)Remove RTs 2SD away from group mean
df3 <- df %>%
group_by(nat_LAN) %>%
filter(!(exp_resp.rt - median(exp_resp.rt)) > abs(2*sd(exp_resp.rt)))
#% of data points removed
1-nrow(df3)/nrow(df)## [1] 0.06933198Visualize RT distribution
ggplot(df3, aes(Idiom_Language, exp_resp.rt, color = Phrase_Type))+
geom_boxplot()+
ylab('Reaction time')+
facet_wrap(~nat_LAN)
Dummy code Binary categorical variables
#preserve df3 as the trimmed dataset with original variable levels
df4<-df3
levels(df4$nat_LAN)<-c(0,1)
colnames(df4)[colnames(df4)=="nat_LAN"] <- "IsBilingual"
levels(df4$Idiom_Language)<-c(1,0)
colnames(df4)[colnames(df4)=="Idiom_Language"] <- "IsChineseIdiom"
levels(df4$Phrase_Type)<-c(0,1)
colnames(df4)[colnames(df4)=="Phrase_Type"] <- "IsExperimental"
df5<-df4[!df4$end=="gold",]
df5<-df5[!df5$end=="water",]
df5<- df5[ -c(17,18) ] #remove duplicate columnsMean and sd of every condition
grouped_mean<-df5 %>%
group_by(IsExperimental, IsBilingual,IsChineseIdiom) %>%
summarise(mean(exp_resp.rt), sd(exp_resp.rt), mean(log(exp_resp.rt)), sd(log(exp_resp.rt)))## `summarise()` regrouping output by 'IsExperimental', 'IsBilingual' (override with `.groups` argument)kable(grouped_mean)| IsExperimental | IsBilingual | IsChineseIdiom | mean(exp_resp.rt) | sd(exp_resp.rt) | mean(log(exp_resp.rt)) | sd(log(exp_resp.rt)) |
|---|---|---|---|---|---|---|
| 0 | 0 | 1 | 636.6360 | 126.6395 | 6.436875 | 0.1962024 |
| 0 | 0 | 0 | 645.6663 | 127.1494 | 6.450971 | 0.1973788 |
| 0 | 1 | 1 | 728.0846 | 169.0169 | 6.564474 | 0.2267021 |
| 0 | 1 | 0 | 729.2199 | 172.1004 | 6.564876 | 0.2322806 |
| 1 | 0 | 1 | 610.0505 | 122.4104 | 6.394016 | 0.1970651 |
| 1 | 0 | 0 | 576.4549 | 138.2126 | 6.328585 | 0.2384026 |
| 1 | 1 | 1 | 702.3808 | 160.9192 | 6.529157 | 0.2240436 |
| 1 | 1 | 0 | 699.2999 | 162.8246 | 6.523844 | 0.2284220 |
Omnibus test
m<-lmer(log(exp_resp.rt)~IsExperimental*IsChineseIdiom*IsBilingual+
length+log(freq_HAL)+
semantic_relatedness+literality+
(1+IsExperimental+IsChineseIdiom|participant)+(1|end),
df5,
control = lmerControl(optimizer = "bobyqa"), REML=FALSE)
summary(m)## Linear mixed model fit by maximum likelihood . t-tests use
## Satterthwaite's method [lmerModLmerTest]
## Formula:
## log(exp_resp.rt) ~ IsExperimental * IsChineseIdiom * IsBilingual +
## length + log(freq_HAL) + semantic_relatedness + literality +
## (1 + IsExperimental + IsChineseIdiom | participant) + (1 | end)
## Data: df5
## Control: lmerControl(optimizer = "bobyqa")
##
## AIC BIC logLik deviance df.resid
## -4428 -4286 2234 -4468 8940
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.9159 -0.6725 -0.0724 0.6181 3.7056
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## end (Intercept) 0.0025510 0.05051
## participant (Intercept) 0.0116526 0.10795
## IsExperimental1 0.0008852 0.02975 0.28
## IsChineseIdiom0 0.0003493 0.01869 -0.25 0.10
## Residual 0.0327306 0.18092
## Number of obs: 8960, groups: end, 170; participant, 122
##
## Fixed effects:
## Estimate Std. Error
## (Intercept) 6.674e+00 5.043e-02
## IsExperimental1 -2.958e-02 1.443e-02
## IsChineseIdiom0 1.250e-02 1.374e-02
## IsBilingual1 1.302e-01 2.111e-02
## length 1.439e-02 3.840e-03
## log(freq_HAL) -2.136e-02 3.294e-03
## semantic_relatedness -3.631e-02 2.634e-02
## literality -2.787e-02 5.822e-03
## IsExperimental1:IsChineseIdiom0 -7.285e-02 1.898e-02
## IsExperimental1:IsBilingual1 8.673e-03 1.231e-02
## IsChineseIdiom0:IsBilingual1 -1.668e-02 1.163e-02
## IsExperimental1:IsChineseIdiom0:IsBilingual1 7.793e-02 1.543e-02
## df t value Pr(>|t|)
## (Intercept) 1.929e+02 132.347 < 2e-16
## IsExperimental1 2.240e+02 -2.049 0.041617
## IsChineseIdiom0 2.235e+02 0.910 0.363607
## IsBilingual1 1.252e+02 6.168 8.81e-09
## length 1.614e+02 3.746 0.000250
## log(freq_HAL) 1.585e+02 -6.487 1.06e-09
## semantic_relatedness 1.829e+02 -1.379 0.169698
## literality 1.770e+02 -4.787 3.56e-06
## IsExperimental1:IsChineseIdiom0 2.160e+02 -3.839 0.000162
## IsExperimental1:IsBilingual1 3.283e+02 0.704 0.481645
## IsChineseIdiom0:IsBilingual1 4.244e+02 -1.434 0.152270
## IsExperimental1:IsChineseIdiom0:IsBilingual1 8.532e+03 5.051 4.48e-07
##
## (Intercept) ***
## IsExperimental1 *
## IsChineseIdiom0
## IsBilingual1 ***
## length ***
## log(freq_HAL) ***
## semantic_relatedness
## literality ***
## IsExperimental1:IsChineseIdiom0 ***
## IsExperimental1:IsBilingual1
## IsChineseIdiom0:IsBilingual1
## IsExperimental1:IsChineseIdiom0:IsBilingual1 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) IsExp1 IsChI0 IsBln1 length l(_HAL smntc_ ltrlty
## IsExprmntl1 -0.016
## IsChinsIdm0 -0.244 0.455
## IsBilingul1 -0.202 0.051 0.130
## length -0.624 -0.043 0.102 0.003
## lg(frq_HAL) -0.793 -0.012 0.070 -0.002 0.271
## smntc_rltdn -0.031 0.016 0.090 0.002 0.032 -0.089
## literality -0.449 -0.231 0.021 0.000 0.214 0.153 -0.169
## IsExp1:ICI0 0.157 -0.646 -0.698 -0.074 -0.039 -0.025 -0.060 -0.084
## IsExpr1:IB1 0.029 -0.401 -0.179 -0.130 -0.006 -0.002 0.000 -0.001
## IsChnI0:IB1 0.062 -0.181 -0.407 -0.323 -0.005 0.004 -0.001 0.004
## IE1:ICI0:IB -0.038 0.257 0.280 0.193 0.006 -0.003 0.000 -0.002
## IsE1:ICI0 IE1:IB ICI0:I
## IsExprmntl1
## IsChinsIdm0
## IsBilingul1
## length
## lg(frq_HAL)
## smntc_rltdn
## literality
## IsExp1:ICI0
## IsExpr1:IB1 0.245
## IsChnI0:IB1 0.269 0.450
## IE1:ICI0:IB -0.385 -0.644 -0.690# reduced model
step_result<-step(m)
final_m <- get_model(step_result)
summary(final_m)## Linear mixed model fit by maximum likelihood . t-tests use
## Satterthwaite's method [lmerModLmerTest]
## Formula:
## log(exp_resp.rt) ~ IsExperimental + IsChineseIdiom + IsBilingual +
## length + log(freq_HAL) + literality + (1 | end) + (IsExperimental |
## participant) + IsExperimental:IsChineseIdiom + IsExperimental:IsBilingual +
## IsChineseIdiom:IsBilingual + IsExperimental:IsChineseIdiom:IsBilingual
## Data: df5
## Control: lmerControl(optimizer = "bobyqa")
##
## AIC BIC logLik deviance df.resid
## -4431.6 -4318.0 2231.8 -4463.6 8944
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.9028 -0.6724 -0.0691 0.6172 3.6916
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## end (Intercept) 0.002529 0.05029
## participant (Intercept) 0.011218 0.10592
## IsExperimental1 0.000881 0.02968 0.29
## Residual 0.032832 0.18120
## Number of obs: 8960, groups: end, 170; participant, 122
##
## Fixed effects:
## Estimate Std. Error
## (Intercept) 6.671e+00 5.019e-02
## IsExperimental1 -2.938e-02 1.439e-02
## IsChineseIdiom0 1.446e-02 1.344e-02
## IsBilingual1 1.301e-01 2.077e-02
## length 1.457e-02 3.826e-03
## log(freq_HAL) -2.175e-02 3.270e-03
## literality -2.919e-02 5.722e-03
## IsExperimental1:IsChineseIdiom0 -7.442e-02 1.889e-02
## IsExperimental1:IsBilingual1 8.926e-03 1.231e-02
## IsChineseIdiom0:IsBilingual1 -1.661e-02 1.113e-02
## IsExperimental1:IsChineseIdiom0:IsBilingual1 7.767e-02 1.545e-02
## df t value Pr(>|t|)
## (Intercept) 1.954e+02 132.917 < 2e-16
## IsExperimental1 2.282e+02 -2.041 0.042364
## IsChineseIdiom0 2.285e+02 1.076 0.282937
## IsBilingual1 1.370e+02 6.265 4.51e-09
## length 1.641e+02 3.809 0.000197
## log(freq_HAL) 1.616e+02 -6.651 4.27e-10
## literality 1.876e+02 -5.102 8.21e-07
## IsExperimental1:IsChineseIdiom0 2.205e+02 -3.939 0.000110
## IsExperimental1:IsBilingual1 3.300e+02 0.725 0.469052
## IsChineseIdiom0:IsBilingual1 8.609e+03 -1.492 0.135818
## IsExperimental1:IsChineseIdiom0:IsBilingual1 8.605e+03 5.028 5.06e-07
##
## (Intercept) ***
## IsExperimental1 *
## IsChineseIdiom0
## IsBilingual1 ***
## length ***
## log(freq_HAL) ***
## literality ***
## IsExperimental1:IsChineseIdiom0 ***
## IsExperimental1:IsBilingual1
## IsChineseIdiom0:IsBilingual1
## IsExperimental1:IsChineseIdiom0:IsBilingual1 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) IsExp1 IsChI0 IsBln1 length l(_HAL ltrlty IsE1:ICI0
## IsExprmntl1 -0.016
## IsChinsIdm0 -0.235 0.458
## IsBilingul1 -0.199 0.051 0.106
## length -0.625 -0.044 0.100 0.003
## lg(frq_HAL) -0.800 -0.010 0.080 -0.002 0.275
## literality -0.462 -0.232 0.038 0.000 0.223 0.141
## IsExp1:ICI0 0.156 -0.647 -0.708 -0.075 -0.037 -0.030 -0.096
## IsExpr1:IB1 0.029 -0.402 -0.178 -0.131 -0.006 -0.002 -0.001 0.246
## IsChnI0:IB1 0.051 -0.184 -0.398 -0.271 -0.005 0.004 0.004 0.283
## IE1:ICI0:IB -0.038 0.258 0.287 0.196 0.006 -0.003 -0.002 -0.387
## IE1:IB ICI0:I
## IsExprmntl1
## IsChinsIdm0
## IsBilingul1
## length
## lg(frq_HAL)
## literality
## IsExp1:ICI0
## IsExpr1:IB1
## IsChnI0:IB1 0.457
## IE1:ICI0:IB -0.644 -0.721significant 3-way interaction
Get native English speaker model
# full model
m_e<-lmer(log(exp_resp.rt)~IsExperimental*IsChineseIdiom+
length+EN_rating+log(freq_HAL)+semantic_relatedness+ literality+
(1+IsExperimental+IsChineseIdiom|participant)+(1|end),
df5[df5$IsBilingual=='0',],
control = lmerControl(optimizer = "bobyqa"), REML=FALSE)
summary(m_e)## Linear mixed model fit by maximum likelihood . t-tests use
## Satterthwaite's method [lmerModLmerTest]
## Formula: log(exp_resp.rt) ~ IsExperimental * IsChineseIdiom + length +
## EN_rating + log(freq_HAL) + semantic_relatedness + literality +
## (1 + IsExperimental + IsChineseIdiom | participant) + (1 | end)
## Data: df5[df5$IsBilingual == "0", ]
## Control: lmerControl(optimizer = "bobyqa")
##
## AIC BIC logLik deviance df.resid
## -2677.3 -2567.6 1355.7 -2711.3 4670
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -6.2931 -0.6424 -0.0489 0.6073 3.6620
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## end (Intercept) 0.0024933 0.04993
## participant (Intercept) 0.0104809 0.10238
## IsExperimental1 0.0013086 0.03618 0.49
## IsChineseIdiom0 0.0006344 0.02519 -0.32 0.07
## Residual 0.0297639 0.17252
## Number of obs: 4687, groups: end, 170; participant, 63
##
## Fixed effects:
## Estimate Std. Error df t value
## (Intercept) 6.664030 0.053826 178.738651 123.808
## IsExperimental1 -0.024858 0.014466 163.639745 -1.718
## IsChineseIdiom0 0.108096 0.052237 157.576153 2.069
## length 0.010737 0.004214 158.964277 2.548
## EN_rating -0.022358 0.011737 155.916628 -1.905
## log(freq_HAL) -0.014498 0.003535 155.725206 -4.101
## semantic_relatedness -0.010441 0.028423 164.847234 -0.367
## literality -0.034564 0.006277 167.482958 -5.506
## IsExperimental1:IsChineseIdiom0 -0.071035 0.018602 153.249022 -3.819
## Pr(>|t|)
## (Intercept) < 2e-16 ***
## IsExperimental1 0.087620 .
## IsChineseIdiom0 0.040147 *
## length 0.011778 *
## EN_rating 0.058628 .
## log(freq_HAL) 6.61e-05 ***
## semantic_relatedness 0.713838
## literality 1.36e-07 ***
## IsExperimental1:IsChineseIdiom0 0.000194 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) IsExp1 IsChI0 length EN_rtn l(_HAL smntc_ ltrlty
## IsExprmntl1 0.023
## IsChinsIdm0 0.048 0.108
## length -0.589 -0.048 0.249
## EN_rating -0.115 0.006 -0.965 -0.230
## lg(frq_HAL) -0.771 -0.014 0.147 0.298 -0.133
## smntc_rltdn -0.038 0.009 0.028 0.035 -0.003 -0.088
## literality -0.455 -0.248 0.044 0.227 -0.040 0.159 -0.148
## IsExp1:ICI0 0.153 -0.626 -0.181 -0.041 0.001 -0.024 -0.067 -0.091# reduced model
step_result<-step(m_e)
final_model <- get_model(step_result)
final_model## Linear mixed model fit by maximum likelihood ['lmerModLmerTest']
## Formula: log(exp_resp.rt) ~ IsExperimental + IsChineseIdiom + length +
## log(freq_HAL) + literality + (1 | end) + (IsExperimental |
## participant) + IsExperimental:IsChineseIdiom
## Data: df5[df5$IsBilingual == "0", ]
## AIC BIC logLik deviance df.resid
## -2679.102 -2601.671 1351.551 -2703.102 4675
## Random effects:
## Groups Name Std.Dev. Corr
## end (Intercept) 0.05059
## participant (Intercept) 0.09894
## IsExperimental1 0.03598 0.51
## Residual 0.17301
## Number of obs: 4687, groups: end, 170; participant, 63
## Fixed Effects:
## (Intercept) IsExperimental1
## 6.651255 -0.024785
## IsChineseIdiom0 length
## 0.012931 0.008993
## log(freq_HAL) literality
## -0.015520 -0.035409
## IsExperimental1:IsChineseIdiom0
## -0.071482summary(final_model)## Linear mixed model fit by maximum likelihood . t-tests use
## Satterthwaite's method [lmerModLmerTest]
## Formula: log(exp_resp.rt) ~ IsExperimental + IsChineseIdiom + length +
## log(freq_HAL) + literality + (1 | end) + (IsExperimental |
## participant) + IsExperimental:IsChineseIdiom
## Data: df5[df5$IsBilingual == "0", ]
## Control: lmerControl(optimizer = "bobyqa")
##
## AIC BIC logLik deviance df.resid
## -2679.1 -2601.7 1351.6 -2703.1 4675
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -6.2772 -0.6462 -0.0418 0.6061 3.6389
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## end (Intercept) 0.002559 0.05059
## participant (Intercept) 0.009790 0.09894
## IsExperimental1 0.001295 0.03598 0.51
## Residual 0.029933 0.17301
## Number of obs: 4687, groups: end, 170; participant, 63
##
## Fixed effects:
## Estimate Std. Error df t value
## (Intercept) 6.651255 0.053813 181.119708 123.599
## IsExperimental1 -0.024785 0.014584 165.909043 -1.699
## IsChineseIdiom0 0.012931 0.013329 160.913824 0.970
## length 0.008993 0.004138 161.412590 2.173
## log(freq_HAL) -0.015520 0.003524 157.912284 -4.404
## literality -0.035409 0.006262 173.562725 -5.655
## IsExperimental1:IsChineseIdiom0 -0.071482 0.018743 155.646622 -3.814
## Pr(>|t|)
## (Intercept) < 2e-16 ***
## IsExperimental1 0.091104 .
## IsChineseIdiom0 0.333425
## length 0.031231 *
## log(freq_HAL) 1.95e-05 ***
## literality 6.31e-08 ***
## IsExperimental1:IsChineseIdiom0 0.000197 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) IsExp1 IsChI0 length l(_HAL ltrlty
## IsExprmntl1 0.024
## IsChinsIdm0 -0.230 0.446
## length -0.638 -0.048 0.107
## lg(frq_HAL) -0.809 -0.012 0.083 0.281
## literality -0.475 -0.250 0.037 0.232 0.144
## IsExp1:ICI0 0.152 -0.628 -0.707 -0.039 -0.030 -0.103No significant main effect of phrase type or phrase language, there is a significant interaction effect
Look at Chinese phrases and English phrases separately
# Chinese phrases
m_ec<-lmer(log(exp_resp.rt)~ IsExperimental+ literality+
length+EN_rating+log(freq_HAL)+
(IsExperimental|participant)+(1|end),
subset(df5, df5$IsChineseIdiom=="1"& df5$IsBilingual=="0"),
control = lmerControl(optimizer = "bobyqa"), REML=FALSE)
summary(m_ec)## Linear mixed model fit by maximum likelihood . t-tests use
## Satterthwaite's method [lmerModLmerTest]
## Formula:
## log(exp_resp.rt) ~ IsExperimental + literality + length + EN_rating +
## log(freq_HAL) + (IsExperimental | participant) + (1 | end)
## Data: subset(df5, df5$IsChineseIdiom == "1" & df5$IsBilingual == "0")
## Control: lmerControl(optimizer = "bobyqa")
##
## AIC BIC logLik deviance df.resid
## -1574.7 -1511.6 798.4 -1596.7 2293
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.5181 -0.6553 -0.0768 0.5951 3.1674
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## end (Intercept) 0.0020849 0.04566
## participant (Intercept) 0.0110869 0.10529
## IsExperimental1 0.0003273 0.01809 0.40
## Residual 0.0259327 0.16104
## Number of obs: 2304, groups: end, 82; participant, 63
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.643222 0.070273 79.495093 94.534 < 2e-16 ***
## IsExperimental1 -0.032935 0.013481 69.821193 -2.443 0.017097 *
## literality -0.021926 0.008942 77.880441 -2.452 0.016447 *
## length 0.005402 0.005438 72.937557 0.993 0.323804
## EN_rating 0.002169 0.016501 72.786692 0.131 0.895769
## log(freq_HAL) -0.016666 0.004439 72.560580 -3.755 0.000347 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) IsExp1 ltrlty length EN_rtn
## IsExprmntl1 0.103
## literality -0.492 -0.391
## length -0.506 -0.037 0.129
## EN_rating -0.334 -0.084 0.210 -0.223
## lg(frq_HAL) -0.768 -0.002 0.087 0.238 0.039Phrase type is now significant, frequency, and literality is significant.
# English phrases
m_ee<-lmer(log(exp_resp.rt)~IsExperimental+literality+
length+EN_rating+log(freq_HAL)+
(IsExperimental|participant)+(1|end),
subset(df5, df5$IsChineseIdiom=="0"& df5$IsBilingual=="0"),
control = lmerControl(optimizer = "bobyqa"), REML=FALSE)
summary(m_ee)## Linear mixed model fit by maximum likelihood . t-tests use
## Satterthwaite's method [lmerModLmerTest]
## Formula:
## log(exp_resp.rt) ~ IsExperimental + literality + length + EN_rating +
## log(freq_HAL) + (IsExperimental | participant) + (1 | end)
## Data: subset(df5, df5$IsChineseIdiom == "0" & df5$IsBilingual == "0")
## Control: lmerControl(optimizer = "bobyqa")
##
## AIC BIC logLik deviance df.resid
## -1037.5 -974.0 529.8 -1059.5 2372
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.7769 -0.6535 -0.0501 0.6372 3.4075
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## end (Intercept) 0.002325 0.04822
## participant (Intercept) 0.008710 0.09333
## IsExperimental1 0.003797 0.06162 0.32
## Residual 0.033207 0.18223
## Number of obs: 2383, groups: end, 88; participant, 63
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.802166 0.098005 82.703538 69.406 < 2e-16 ***
## IsExperimental1 -0.091286 0.016785 97.600697 -5.439 3.97e-07 ***
## literality -0.038177 0.008566 88.389842 -4.457 2.43e-05 ***
## length 0.013936 0.006070 82.012900 2.296 0.0242 *
## EN_rating -0.036472 0.016703 79.431723 -2.184 0.0319 *
## log(freq_HAL) -0.010435 0.005341 79.628824 -1.954 0.0542 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) IsExp1 ltrlty length EN_rtn
## IsExprmntl1 0.068
## literality -0.210 -0.449
## length -0.288 -0.153 0.320
## EN_rating -0.657 0.105 -0.257 -0.280
## lg(frq_HAL) -0.396 -0.114 0.275 0.366 -0.318Phrase type is still significant. Familarity rating, literality, word length are significant. Frequency marginally significant, showing English participants’ RT was not affected by word frequency.
Get bilingual particpants model
# full model
m_c<-lmer(log(exp_resp.rt)~IsExperimental*IsChineseIdiom+length+ CN_rating+ literality+
semantic_relatedness+log(freq_HAL)+CAN_year+
(IsExperimental+IsChineseIdiom|participant)+(1|end),
df5[df5$IsBilingual=='1',],
control = lmerControl(optimizer = "bobyqa"), REML=FALSE)
summary(m_c)## Linear mixed model fit by maximum likelihood . t-tests use
## Satterthwaite's method [lmerModLmerTest]
## Formula: log(exp_resp.rt) ~ IsExperimental * IsChineseIdiom + length +
## CN_rating + literality + semantic_relatedness + log(freq_HAL) +
## CAN_year + (IsExperimental + IsChineseIdiom | participant) +
## (1 | end)
## Data: df5[df5$IsBilingual == "1", ]
## Control: lmerControl(optimizer = "bobyqa")
##
## AIC BIC logLik deviance df.resid
## -1800.7 -1686.2 918.4 -1836.7 4255
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.7175 -0.6747 -0.1050 0.6127 3.5911
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## end (Intercept) 0.0034385 0.05864
## participant (Intercept) 0.0117841 0.10855
## IsExperimental1 0.0003869 0.01967 0.05
## IsChineseIdiom0 0.0002289 0.01513 0.01 -0.08
## Residual 0.0345525 0.18588
## Number of obs: 4273, groups: end, 170; participant, 59
##
## Fixed effects:
## Estimate Std. Error df t value
## (Intercept) 6.883e+00 7.601e-02 2.339e+02 90.553
## IsExperimental1 -2.832e-02 1.627e-02 1.574e+02 -1.741
## IsChineseIdiom0 -4.853e-03 1.970e-02 1.732e+02 -0.246
## length 2.192e-02 4.759e-03 1.649e+02 4.605
## CN_rating -6.279e-04 6.338e-03 1.990e+02 -0.099
## literality -1.967e-02 7.238e-03 1.727e+02 -2.717
## semantic_relatedness -4.718e-02 3.279e-02 1.740e+02 -1.439
## log(freq_HAL) -2.788e-02 4.086e-03 1.622e+02 -6.822
## CAN_year -1.255e-02 6.717e-03 5.882e+01 -1.868
## IsExperimental1:IsChineseIdiom0 6.594e-03 2.167e-02 1.590e+02 0.304
## Pr(>|t|)
## (Intercept) < 2e-16 ***
## IsExperimental1 0.08371 .
## IsChineseIdiom0 0.80566
## length 8.21e-06 ***
## CN_rating 0.92118
## literality 0.00725 **
## semantic_relatedness 0.15208
## log(freq_HAL) 1.69e-10 ***
## CAN_year 0.06671 .
## IsExperimental1:IsChineseIdiom0 0.76126
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) IsExp1 IsChI0 length CN_rtn ltrlty smntc_ l(_HAL CAN_yr
## IsExprmntl1 -0.010
## IsChinsIdm0 -0.376 0.363
## length -0.496 -0.056 0.062
## CN_rating -0.362 0.005 0.617 -0.031
## literality -0.372 -0.259 0.023 0.211 0.001
## smntc_rltdn -0.054 0.011 0.111 0.036 0.056 -0.152
## lg(frq_HAL) -0.664 -0.018 0.082 0.261 0.030 0.158 -0.084
## CAN_year -0.459 0.005 0.002 0.002 0.000 -0.005 0.004 0.002
## IsExp1:ICI0 0.135 -0.655 -0.563 -0.032 -0.020 -0.095 -0.067 -0.029 -0.003# reduced model
step_resultc<-step(m_c)
final_model_c <- get_model(step_resultc)
#final_model_c
summary(final_model_c)## Linear mixed model fit by maximum likelihood . t-tests use
## Satterthwaite's method [lmerModLmerTest]
## Formula:
## log(exp_resp.rt) ~ IsExperimental + length + literality + log(freq_HAL) +
## (1 | end) + (1 | participant)
## Data: df5[df5$IsBilingual == "1", ]
## Control: lmerControl(optimizer = "bobyqa")
##
## AIC BIC logLik deviance df.resid
## -1813.8 -1762.9 914.9 -1829.8 4265
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.8145 -0.6780 -0.1049 0.6126 3.5662
##
## Random effects:
## Groups Name Variance Std.Dev.
## end (Intercept) 0.003436 0.05862
## participant (Intercept) 0.012783 0.11306
## Residual 0.034726 0.18635
## Number of obs: 4273, groups: end, 170; participant, 59
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.810274 0.059998 189.958830 113.509 < 2e-16 ***
## IsExperimental1 -0.026235 0.012018 164.367855 -2.183 0.03045 *
## length 0.022136 0.004721 166.853395 4.689 5.69e-06 ***
## literality -0.021056 0.007107 179.100732 -2.963 0.00346 **
## log(freq_HAL) -0.028345 0.004050 164.419871 -6.999 6.23e-11 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) IsExp1 length ltrlty
## IsExprmntl1 0.129
## length -0.619 -0.105
## literality -0.479 -0.449 0.224
## lg(frq_HAL) -0.814 -0.055 0.257 0.149Main effect of phrase type, no significant interaction effect.
Look at Chinese phrases and English phrases separately
# Chinese phrases
m_cc<-lmer(log(exp_resp.rt)~ IsExperimental+literality+
length+log(freq_HAL)+
(1|participant)+(1|end),
subset(df5, df5$IsChineseIdiom=="1"& df5$IsBilingual=="1"),
control = lmerControl(optimizer = "bobyqa"), REML=FALSE)
summary(m_cc)## Linear mixed model fit by maximum likelihood . t-tests use
## Satterthwaite's method [lmerModLmerTest]
## Formula:
## log(exp_resp.rt) ~ IsExperimental + literality + length + log(freq_HAL) +
## (1 | participant) + (1 | end)
## Data: subset(df5, df5$IsChineseIdiom == "1" & df5$IsBilingual == "1")
## Control: lmerControl(optimizer = "bobyqa")
##
## AIC BIC logLik deviance df.resid
## -850.8 -805.8 433.4 -866.8 2052
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.5256 -0.6896 -0.0888 0.6108 3.5743
##
## Random effects:
## Groups Name Variance Std.Dev.
## end (Intercept) 0.00355 0.05958
## participant (Intercept) 0.01287 0.11344
## Residual 0.03392 0.18417
## Number of obs: 2060, groups: end, 82; participant, 59
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.744956 0.084201 85.557403 80.105 < 2e-16 ***
## IsExperimental1 -0.036412 0.017008 79.670097 -2.141 0.03535 *
## literality -0.006694 0.011123 85.604300 -0.602 0.54887
## length 0.021793 0.006804 80.389343 3.203 0.00195 **
## log(freq_HAL) -0.025221 0.005708 79.082626 -4.419 3.12e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) IsExp1 ltrlty length
## IsExprmntl1 0.076
## literality -0.459 -0.391
## length -0.621 -0.068 0.176
## lg(frq_HAL) -0.807 -0.006 0.085 0.237Phrase type is significant, frequency and length are significant.
# English phrases
m_ce<-lmer(log(exp_resp.rt)~IsExperimental+literality+
length+log(freq_HAL)+
(1|participant)+(1|end),
subset(df5, df5$IsChineseIdiom=="0"& df5$IsBilingual=="1"),
control = lmerControl(optimizer = "bobyqa"), REML=FALSE)
summary(m_ce)## Linear mixed model fit by maximum likelihood . t-tests use
## Satterthwaite's method [lmerModLmerTest]
## Formula:
## log(exp_resp.rt) ~ IsExperimental + literality + length + log(freq_HAL) +
## (1 | participant) + (1 | end)
## Data: subset(df5, df5$IsChineseIdiom == "0" & df5$IsBilingual == "1")
## Control: lmerControl(optimizer = "bobyqa")
##
## AIC BIC logLik deviance df.resid
## -842.8 -797.2 429.4 -858.8 2205
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.8511 -0.6636 -0.0927 0.5863 3.2372
##
## Random effects:
## Groups Name Variance Std.Dev.
## end (Intercept) 0.003246 0.05698
## participant (Intercept) 0.012944 0.11377
## Residual 0.035314 0.18792
## Number of obs: 2213, groups: end, 88; participant, 59
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.881275 0.083735 92.668491 82.179 < 2e-16 ***
## IsExperimental1 -0.012823 0.016944 83.230434 -0.757 0.451307
## literality -0.032559 0.009320 91.688111 -3.493 0.000736 ***
## length 0.021850 0.006597 83.931689 3.312 0.001367 **
## log(freq_HAL) -0.032659 0.005780 83.182472 -5.650 2.19e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) IsExp1 ltrlty length
## IsExprmntl1 0.188
## literality -0.510 -0.503
## length -0.644 -0.141 0.259
## lg(frq_HAL) -0.846 -0.109 0.205 0.297Phrase type no longer significant, frequency, literality and length are significant.