Experiment 1
- 1 Analyses for Experiment 1
- 1.1 Priming analyses for all targets together
- 1.2 Analyses with target stress position interacting with condition
- 1.3 Graphs based on the interaction models
- 1.4 Target stress position interacting with condition, match vs. mismatch only (i.e., excluding Neutral)
- 1.5 Models for pairwise comparisons of priming effects
- 1.6 Bayes factor for the interaction between stress position and stress priming
- 1.7 Session information
1 Analyses for Experiment 1
This file presents the analyses reported in the manuscript titled “Lexical stress representation in Greek spoken word recognition” by Andrikopoulou et al. See the file “mega script.R” for pre-processing and additional analyses.
Note: Models with more complex random structures either fail to converge or result in singular fits. We have tried all available optimizers with allFit and confirmed overparameterization in singular fits with rePCA.
read.table("alldata15.txt",stringsAsFactors = TRUE) -> data3
data3$item <- as.factor(data3$item)
data3$strpos <- as.factor(data3$strpos)
data3$targpos <- relevel(data3$targpos,ref="Pen")
data3$cond <- relevel(data3$cond,ref="Neutral")
data3$CprevRT <- scale(log(data3$prevRT)) # center to facilitate convergence
data3$Corder <- scale(data3$order) # center to facilitate convergence
data3$condC <- ifelse(data3$cond=="Match",-0.5,ifelse(data3$cond=="Mismatch",+0.5,NA))
str(data3)## 'data.frame': 21312 obs. of 23 variables:
## $ name : Factor w/ 288 levels "á÷áøß","äåäáìÜò",..: 15 15 15 15 15 15 15 15 15 15 ...
## $ item : Factor w/ 864 levels "210147","210150",..: 313 313 313 313 313 73 313 313 313 73 ...
## $ subject : Factor w/ 74 levels "aggi","agka",..: 59 60 2 11 58 54 37 68 67 46 ...
## $ order : int 265 60 78 295 155 88 161 59 134 288 ...
## $ RT : num 512 890 1043 586 1056 ...
## $ prevRT : num 511 684 1260 633 1072 ...
## $ lex : Factor w/ 2 levels "Pseudo","Word": 1 1 1 1 1 1 1 1 1 1 ...
## $ list : Factor w/ 3 levels "L1","L2","L3": 3 3 3 3 3 1 3 3 3 1 ...
## $ cond : Factor w/ 3 levels "Neutral","Match",..: 1 1 1 1 1 2 1 1 1 2 ...
## $ strpos : Factor w/ 10 levels "0","11","12",..: 1 1 1 1 1 6 1 1 1 6 ...
## $ logprevRT: num 6.24 6.53 7.14 6.45 6.98 ...
## $ targstr : Factor w/ 3 levels "T1","T2","T3": NA NA NA NA NA 2 NA NA NA 2 ...
## $ primstr : Factor w/ 3 levels "P1","P2","P3": NA NA NA NA NA 2 NA NA NA 2 ...
## $ err : int 0 0 0 0 0 0 0 0 0 0 ...
## $ cRT : num 512 890 1043 586 1056 ...
## $ logRT : num 6.24 6.79 6.95 6.37 6.96 ...
## $ CprevRT : num [1:21312, 1] -1.19 -0.241 1.743 -0.495 1.218 ...
## ..- attr(*, "scaled:center")= num 6.6
## ..- attr(*, "scaled:scale")= num 0.308
## $ Corder : num [1:21312, 1] 1.092 -1.226 -1.022 1.431 -0.152 ...
## ..- attr(*, "scaled:center")= num 168
## ..- attr(*, "scaled:scale")= num 88.5
## $ Fin : logi FALSE FALSE FALSE FALSE FALSE FALSE ...
## $ Pen : logi TRUE TRUE TRUE TRUE TRUE TRUE ...
## $ Ant : logi FALSE FALSE FALSE FALSE FALSE FALSE ...
## $ targpos : Factor w/ 3 levels "Pen","Ant","Fin": 1 1 1 1 1 1 1 1 1 1 ...
## $ condC : num NA NA NA NA NA -0.5 NA NA NA -0.5 ...1.1 Priming analyses for all targets together
Treatment coded; reference level is Neutral.
1.1.1 Words
1.1.1.1 Model for latency (RT)
wrt <- lmer ( logRT ~ cond + CprevRT + Corder + (1+Corder|subject) + (cond|name), data3, subset= lex=="Word" , control = lmerControl(optimizer = "bobyqa"))
summary(wrt)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ cond + CprevRT + Corder + (1 + Corder | subject) + (cond |
## name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Word"
##
## REML criterion at convergence: -1685.8
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.3693 -0.6403 -0.0998 0.5239 6.0891
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## name (Intercept) 0.016128 0.12699
## condMatch 0.004831 0.06950 -0.10
## condMismatch 0.002668 0.05165 -0.23 0.39
## subject (Intercept) 0.029625 0.17212
## Corder 0.001942 0.04407 0.26
## Residual 0.043322 0.20814
## Number of obs: 9413, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.595e+00 2.295e-02 1.158e+02 287.391 < 2e-16 ***
## condMatch -5.904e-02 7.843e-03 1.347e+02 -7.528 6.64e-12 ***
## condMismatch 5.090e-02 6.875e-03 1.318e+02 7.404 1.40e-11 ***
## CprevRT 2.535e-02 2.690e-03 9.198e+03 9.427 < 2e-16 ***
## Corder -9.220e-03 5.583e-03 7.261e+01 -1.651 0.103
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) cndMtc cndMsm CprvRT
## condMatch -0.114
## condMismtch -0.154 0.441
## CprevRT -0.001 0.002 -0.004
## Corder 0.212 -0.001 -0.003 0.0211.1.1.2 Model for accuracy (errors)
wer<-glmer(err~cond+(1|subject)+(1|name), data = data3, family = binomial, subset= lex=="Word")
summary(wer)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: err ~ cond + (1 | subject) + (1 | name)
## Data: data3
## Subset: lex == "Word"
##
## AIC BIC logLik deviance df.resid
## 6089.7 6126.1 -3039.9 6079.7 10651
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.0909 -0.3345 -0.1763 -0.0989 13.0705
##
## Random effects:
## Groups Name Variance Std.Dev.
## name (Intercept) 2.7369 1.6543
## subject (Intercept) 0.3752 0.6125
## Number of obs: 10656, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.99519 0.17388 -17.226 < 2e-16 ***
## condMatch -0.30004 0.08622 -3.480 0.000501 ***
## condMismatch 0.30098 0.07980 3.771 0.000162 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) cndMtc
## condMatch -0.220
## condMismtch -0.254 0.4941.1.2 Pseudowords
1.1.2.1 Model for latency (RT)
prt <- lmer ( logRT ~ cond + CprevRT + Corder + (1+Corder|subject) + (cond|name), data3, subset= lex=="Pseudo" , control = lmerControl(optimizer = "bobyqa"))
summary(prt)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ cond + CprevRT + Corder + (1 + Corder | subject) + (cond |
## name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo"
##
## REML criterion at convergence: -5258
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.5709 -0.6368 -0.1108 0.5053 5.6421
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## name (Intercept) 0.008479 0.09208
## condMatch 0.002704 0.05200 -0.68
## condMismatch 0.001814 0.04259 -0.68 0.85
## subject (Intercept) 0.023942 0.15473
## Corder 0.001822 0.04269 0.16
## Residual 0.031782 0.17828
## Number of obs: 10292, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.625e+00 1.979e-02 1.018e+02 334.670 < 2e-16 ***
## condMatch -4.866e-02 6.124e-03 1.344e+02 -7.946 6.90e-13 ***
## condMismatch -2.818e-02 5.598e-03 1.300e+02 -5.035 1.56e-06 ***
## CprevRT 3.215e-02 2.222e-03 1.009e+04 14.468 < 2e-16 ***
## Corder -1.494e-02 5.270e-03 7.248e+01 -2.835 0.00593 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) cndMtc cndMsm CprvRT
## condMatch -0.264
## condMismtch -0.253 0.655
## CprevRT 0.002 -0.001 -0.013
## Corder 0.136 0.004 0.004 0.0191.1.2.2 Model for accuracy (errors)
per<-glmer(err~cond+(1|subject)+(1|name), data = data3, family = binomial, subset= lex=="Pseudo")
summary(per)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: err ~ cond + (1 | subject) + (1 | name)
## Data: data3
## Subset: lex == "Pseudo"
##
## AIC BIC logLik deviance df.resid
## 2644.2 2680.5 -1317.1 2634.2 10651
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.6182 -0.1661 -0.1183 -0.0857 13.1541
##
## Random effects:
## Groups Name Variance Std.Dev.
## name (Intercept) 1.4344 1.1977
## subject (Intercept) 0.4837 0.6955
## Number of obs: 10656, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.18331 0.18180 -23.011 <2e-16 ***
## condMatch -0.15728 0.14534 -1.082 0.279
## condMismatch -0.09169 0.14265 -0.643 0.520
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) cndMtc
## condMatch -0.363
## condMismtch -0.366 0.4791.1.3 Compare words to pseudowords
1.1.3.1 Model for latency (RT)
wprt <- lmer ( logRT ~ cond*lex + CprevRT + Corder + (1+Corder|subject) + (cond|name), data3, control = lmerControl(optimizer = "bobyqa"))
print(summary(wprt),corr=F) ## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ cond * lex + CprevRT + Corder + (1 + Corder | subject) +
## (cond | name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
##
## REML criterion at convergence: -6019.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.6511 -0.6421 -0.0990 0.5143 6.5115
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## name (Intercept) 0.012129 0.11013
## condMatch 0.003728 0.06106 -0.34
## condMismatch 0.002146 0.04633 -0.41 0.57
## subject (Intercept) 0.024802 0.15749
## Corder 0.001676 0.04094 0.23
## Residual 0.039395 0.19848
## Number of obs: 19705, groups: name, 288; subject, 74
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.624e+00 2.076e-02 1.160e+02 319.054 < 2e-16 ***
## condMatch -4.859e-02 7.004e-03 2.577e+02 -6.937 3.22e-11 ***
## condMismatch -2.789e-02 6.172e-03 2.482e+02 -4.520 9.60e-06 ***
## lexWord -3.048e-02 1.389e-02 2.784e+02 -2.194 0.0291 *
## CprevRT 2.788e-02 1.778e-03 1.930e+04 15.683 < 2e-16 ***
## Corder -1.272e-02 4.972e-03 7.258e+01 -2.559 0.0126 *
## condMatch:lexWord -1.039e-02 1.001e-02 2.673e+02 -1.038 0.3000
## condMismatch:lexWord 7.857e-02 8.885e-03 2.635e+02 8.843 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 11.1.3.2 Model for accuracy (errors)
wper <- glmer( err ~ cond*lex + (1|subject) + (1|name), data = data3, family = binomial, control = glmerControl(optimizer = "bobyqa"))
print(summary(wper),corr=F)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: err ~ cond * lex + (1 | subject) + (1 | name)
## Data: data3
## Control: glmerControl(optimizer = "bobyqa")
##
## AIC BIC logLik deviance df.resid
## 8813.0 8876.7 -4398.5 8797.0 21304
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.2558 -0.2330 -0.1414 -0.0939 13.3452
##
## Random effects:
## Groups Name Variance Std.Dev.
## name (Intercept) 2.0545 1.4334
## subject (Intercept) 0.2445 0.4944
## Number of obs: 21312, groups: name, 288; subject, 74
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.18670 0.17211 -24.326 < 2e-16 ***
## condMatch -0.12040 0.13513 -0.891 0.3729
## condMismatch -0.05552 0.13325 -0.417 0.6769
## lexWord 1.30083 0.21039 6.183 6.29e-10 ***
## condMatch:lexWord -0.17806 0.15996 -1.113 0.2657
## condMismatch:lexWord 0.35208 0.15496 2.272 0.0231 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 11.2 Analyses with target stress position interacting with condition
Treatment-coded; reference levels are penultimate-syllable-stress targets (for targpos) and Neutral (for cond).
1.2.1 Words
1.2.1.1 Model for latency
lint <- lmer (logRT ~ targpos*cond + CprevRT + Corder + (1+Corder|subject) + (cond|name), data3, subset=lex=="Word", control = lmerControl(optimizer = "bobyqa"))
slint <- summary(lint)
print(slint,corr=F)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ targpos * cond + CprevRT + Corder + (1 + Corder | subject) +
## (cond | name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Word"
##
## REML criterion at convergence: -1668.9
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.3976 -0.6351 -0.1034 0.5215 6.0716
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## name (Intercept) 0.014997 0.12246
## condMatch 0.004707 0.06861 -0.07
## condMismatch 0.002596 0.05095 -0.26 0.39
## subject (Intercept) 0.029657 0.17221
## Corder 0.001941 0.04406 0.26
## Residual 0.043325 0.20815
## Number of obs: 9413, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.611e+00 2.751e-02 1.782e+02 240.318 < 2e-16 ***
## targposAnt 1.959e-02 2.669e-02 1.374e+02 0.734 0.46428
## targposFin -6.680e-02 2.662e-02 1.360e+02 -2.509 0.01327 *
## condMatch -7.979e-02 1.357e-02 1.356e+02 -5.881 3.02e-08 ***
## condMismatch 3.728e-02 1.190e-02 1.318e+02 3.134 0.00213 **
## CprevRT 2.531e-02 2.690e-03 9.200e+03 9.412 < 2e-16 ***
## Corder -9.182e-03 5.582e-03 7.261e+01 -1.645 0.10430
## targposAnt:condMatch 2.315e-02 1.920e-02 1.359e+02 1.206 0.22993
## targposFin:condMatch 3.878e-02 1.901e-02 1.313e+02 2.040 0.04331 *
## targposAnt:condMismatch 2.905e-02 1.692e-02 1.338e+02 1.716 0.08845 .
## targposFin:condMismatch 1.201e-02 1.662e-02 1.266e+02 0.723 0.47118
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 11.2.1.2 Model for accuracy
acint <- glmer (err ~ targpos*cond + (1|subject) + (1|name), data3, family = binomial, subset=lex=="Word", control = glmerControl(optimizer = "bobyqa"))
acints <- summary(acint)
print(acints,corr=F)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: err ~ targpos * cond + (1 | subject) + (1 | name)
## Data: data3
## Control: glmerControl(optimizer = "bobyqa")
## Subset: lex == "Word"
##
## AIC BIC logLik deviance df.resid
## 6084.2 6164.2 -3031.1 6062.2 10645
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.1383 -0.3344 -0.1762 -0.0977 14.7773
##
## Random effects:
## Groups Name Variance Std.Dev.
## name (Intercept) 2.5196 1.5873
## subject (Intercept) 0.3755 0.6128
## Number of obs: 10656, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.756603 0.267739 -10.296 <2e-16 ***
## targposAnt 0.140320 0.360360 0.389 0.6970
## targposFin -0.827101 0.373110 -2.217 0.0266 *
## condMatch -0.249606 0.138694 -1.800 0.0719 .
## condMismatch 0.124218 0.132218 0.939 0.3475
## targposAnt:condMatch -0.002825 0.195546 -0.014 0.9885
## targposFin:condMatch -0.224981 0.230473 -0.976 0.3290
## targposAnt:condMismatch 0.384459 0.183245 2.098 0.0359 *
## targposFin:condMismatch 0.097453 0.210594 0.463 0.6435
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 11.2.2 Pseudowords
1.2.2.1 Model for latency
pint <- lmer (logRT ~ targpos*cond + CprevRT + Corder + (1+Corder|subject) + (cond|name), data3, subset=lex=="Pseudo", control = lmerControl(optimizer = "bobyqa"))
spint <- summary(pint)
print(spint,corr=F)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ targpos * cond + CprevRT + Corder + (1 + Corder | subject) +
## (cond | name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo"
##
## REML criterion at convergence: -5235.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.5965 -0.6375 -0.1099 0.5047 5.6421
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## name (Intercept) 0.008051 0.08973
## condMatch 0.002755 0.05249 -0.72
## condMismatch 0.001815 0.04261 -0.69 0.85
## subject (Intercept) 0.023942 0.15473
## Corder 0.001827 0.04274 0.16
## Residual 0.031780 0.17827
## Number of obs: 10292, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.621e+00 2.279e-02 1.541e+02 290.557 < 2e-16 ***
## targposAnt -2.369e-02 1.980e-02 1.391e+02 -1.197 0.233517
## targposFin 3.382e-02 1.979e-02 1.391e+02 1.709 0.089735 .
## condMatch -4.230e-02 1.063e-02 1.318e+02 -3.978 0.000114 ***
## condMismatch -1.634e-02 9.702e-03 1.290e+02 -1.684 0.094603 .
## CprevRT 3.209e-02 2.222e-03 1.009e+04 14.440 < 2e-16 ***
## Corder -1.499e-02 5.276e-03 7.248e+01 -2.842 0.005826 **
## targposAnt:condMatch -1.206e-02 1.507e-02 1.324e+02 -0.800 0.425181
## targposFin:condMatch -7.286e-03 1.506e-02 1.325e+02 -0.484 0.629286
## targposAnt:condMismatch -1.306e-02 1.373e-02 1.287e+02 -0.951 0.343259
## targposFin:condMismatch -2.250e-02 1.370e-02 1.286e+02 -1.642 0.103048
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 11.2.2.2 Model for accuracy
acpsint <- glmer (err ~ targpos*cond + (1|subject) + (1|name), data3, family = binomial, subset=lex=="Pseudo", control = glmerControl(optimizer = "bobyqa"))
acpsints <- summary(acpsint)
print(acpsints,corr=F)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: err ~ targpos * cond + (1 | subject) + (1 | name)
## Data: data3
## Control: glmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo"
##
## AIC BIC logLik deviance df.resid
## 2640.8 2720.8 -1309.4 2618.8 10645
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.4076 -0.1664 -0.1156 -0.0841 12.3395
##
## Random effects:
## Groups Name Variance Std.Dev.
## name (Intercept) 1.4449 1.2020
## subject (Intercept) 0.4881 0.6987
## Number of obs: 10656, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.2599 0.2811 -15.154 <2e-16 ***
## targposAnt -0.1419 0.3730 -0.380 0.7036
## targposFin 0.3223 0.3578 0.901 0.3678
## condMatch -0.5418 0.2772 -1.955 0.0506 .
## condMismatch 0.2171 0.2362 0.919 0.3579
## targposAnt:condMatch 0.7156 0.3739 1.914 0.0556 .
## targposFin:condMatch 0.3791 0.3605 1.052 0.2929
## targposAnt:condMismatch -0.1034 0.3447 -0.300 0.7642
## targposFin:condMismatch -0.8843 0.3515 -2.516 0.0119 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 11.3 Graphs based on the interaction models
errorbars <- function(d) {
x <- d[1]
y <- d[2]
lo <- d[3]
hi <- d[4]
arrows(x, lo, x, hi, length=0.05, angle=90, code=3)
}
dograph <- function(eff,mtxt,ytxt,ylm=NULL,tps=NULL,FUN=identity) {
effw<-data.frame(cond=rep(eff$variables$cond$levels,3),
targpos=rep(eff$variables$targpos$levels,each=3),
fit=eff$fit, lower=eff$lower, upper=eff$upper)
effw$condF<-factor(effw$cond,ordered=T,levels=c("Match","Neutral","Mismatch"))
effw$cond<-as.numeric(effw$condF)
effw$targpos<-factor(effw$targpos,ordered=T,levels=c("Ant","Pen","Fin"))
effw$fitms<-FUN(effw$fit)
effw$loms<-FUN(effw$lower)
effw$hims<-FUN(effw$upper)
effw$x <- effw$cond+0.01*(as.numeric(effw$targpos)-2)
effw <- effw[order(effw$targpos,effw$cond),]
plot(1:3,c(min(effw$loms),median(effw$fitms),max(effw$hims)),type="n",las=1,xlab="Target stress position",ylab=ytxt,xaxt="n",main=mtxt,ylim=ylm)
axis(1,at=1:3,labels=levels(effw$condF))
for (i in 1:3) {
tp <- levels(effw$targpos)[i]
points(fitms~x,effw,subset= targpos==tp)
lines(fitms~x,effw,subset= targpos==tp)
apply(effw[effw$targpos==tp,c("x","fitms","loms","hims")],1,errorbars)
text(2.9,effw$fitms[effw$targpos==tp & effw$cond==3],tp,pos=tps) # try pos=3 for pseudowords
}
}par(mfrow=c(2,2))
par(oma=c(0,1.5,0,0))
par(mai=c(0.3,0.4,0.3,0.05))
par(mgp=c(2,0.5,0))
par(tcl=-0.35)
dograph(Effect(c("cond","targpos"),lint),FUN=exp,mtxt="Words",ytxt="",ylm=c(630,860),tps=3)
mtext("Response time (ms)",2,line=2.75)
dograph(Effect(c("cond","targpos"),pint),FUN=exp,mtxt="Pseudowords",ytxt="",ylm=c(630,860),tps=3)
dograph(Effect(c("cond","targpos"),acint),FUN=inv.logit,mtxt="Words",ytxt="",ylm=c(0,0.20),tps=3)
mtext("Error proportion",2,line=2.75)
dograph(Effect(c("cond","targpos"),acpsint),FUN=inv.logit,mtxt="Pseudowords",ytxt="",ylm=c(0,0.20),tps=3)
These are model-based estimates and associated 95% confidence intervals (back-transformed to ms and proportion).
1.4 Target stress position interacting with condition, match vs. mismatch only (i.e., excluding Neutral)
Treatment-coded target stress position; reference is penultimate. Difference-coded condition.
1.4.1 Words
1.4.1.1 Models for latency
1.4.1.1.1 All targets together
lrtn <- lmer ( logRT ~ cond + CprevRT + Corder + (1+Corder|subject) + (cond|name), data3, subset= lex=="Word" & cond!="Neutral", control = lmerControl(optimizer = "bobyqa"))
summary(lrtn) # ## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ cond + CprevRT + Corder + (1 + Corder | subject) + (cond |
## name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Word" & cond != "Neutral"
##
## REML criterion at convergence: -818.7
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.2160 -0.6366 -0.0928 0.5270 5.9071
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## name (Intercept) 0.019139 0.13834
## condMismatch 0.004668 0.06832 -0.43
## subject (Intercept) 0.029802 0.17263
## Corder 0.001790 0.04231 0.19
## Residual 0.044232 0.21031
## Number of obs: 6271, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.536e+00 2.345e-02 1.202e+02 278.708 < 2e-16 ***
## condMismatch 1.096e-01 7.850e-03 1.306e+02 13.966 < 2e-16 ***
## CprevRT 2.774e-02 3.416e-03 6.111e+03 8.122 5.51e-16 ***
## Corder -8.001e-03 5.642e-03 7.106e+01 -1.418 0.161
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) cndMsm CprvRT
## condMismtch -0.228
## CprevRT 0.000 -0.007
## Corder 0.138 -0.003 0.0261.4.1.1.2 By target stress position
lintn <- lmer (logRT ~ targpos*condC + CprevRT + Corder + (1+Corder|subject) + (cond|name), data3, subset= lex=="Word" & cond!="Neutral", control = lmerControl(optimizer = "bobyqa"))
slintn <- summary(lintn)
print(slintn,corr=F)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ targpos * condC + CprevRT + Corder + (1 + Corder | subject) +
## (cond | name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Word" & cond != "Neutral"
##
## REML criterion at convergence: -810.9
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.2279 -0.6364 -0.0922 0.5263 5.8960
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## name (Intercept) 0.018619 0.13645
## condMismatch 0.004535 0.06734 -0.48
## subject (Intercept) 0.029882 0.17286
## Corder 0.001792 0.04233 0.19
## Residual 0.044234 0.21032
## Number of obs: 6271, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.589e+00 2.730e-02 1.733e+02 241.335 < 2e-16 ***
## targposAnt 4.491e-02 2.616e-02 1.386e+02 1.717 0.0882 .
## targposFin -4.132e-02 2.609e-02 1.372e+02 -1.584 0.1156
## condC 1.166e-01 1.355e-02 1.292e+02 8.604 2.24e-14 ***
## CprevRT 2.765e-02 3.415e-03 6.112e+03 8.095 6.88e-16 ***
## Corder -7.945e-03 5.644e-03 7.107e+01 -1.408 0.1636
## targposAnt:condC 5.592e-03 1.924e-02 1.310e+02 0.291 0.7718
## targposFin:condC -2.593e-02 1.895e-02 1.249e+02 -1.368 0.1737
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 11.4.1.2 Models for accuracy
1.4.1.2.1 All targets together
wern<-glmer(err~condC+(1|subject)+(1|name), data = data3, family = binomial, subset= lex=="Word" & cond!="Neutral", control = glmerControl(optimizer = "bobyqa"))
summary(wern)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: err ~ condC + (1 | subject) + (1 | name)
## Data: data3
## Control: glmerControl(optimizer = "bobyqa")
## Subset: lex == "Word" & cond != "Neutral"
##
## AIC BIC logLik deviance df.resid
## 4096.6 4124.1 -2044.3 4088.6 7100
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.3131 -0.3209 -0.1768 -0.1058 10.7586
##
## Random effects:
## Groups Name Variance Std.Dev.
## name (Intercept) 2.7091 1.6459
## subject (Intercept) 0.3419 0.5847
## Number of obs: 7104, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.99092 0.16983 -17.611 < 2e-16 ***
## condC 0.60965 0.08393 7.264 3.76e-13 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## condC -0.0631.4.1.2.2 By target stress position
acintn <- glmer (err ~ targpos*condC + (1|subject) + (1|name), data3, family = binomial, subset= lex=="Word" & cond!="Neutral", control = glmerControl(optimizer = "bobyqa"))
acintsn <- summary(acintn)
print(acintsn,corr=F)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: err ~ targpos * condC + (1 | subject) + (1 | name)
## Data: data3
## Control: glmerControl(optimizer = "bobyqa")
## Subset: lex == "Word" & cond != "Neutral"
##
## AIC BIC logLik deviance df.resid
## 4090.2 4145.1 -2037.1 4074.2 7096
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.1706 -0.3190 -0.1780 -0.1046 11.8927
##
## Random effects:
## Groups Name Variance Std.Dev.
## name (Intercept) 2.4991 1.5808
## subject (Intercept) 0.3414 0.5843
## Number of obs: 7104, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.8620 0.2623 -10.910 < 2e-16 ***
## targposAnt 0.3881 0.3494 1.111 0.26674
## targposFin -0.7888 0.3607 -2.187 0.02875 *
## condC 0.3949 0.1385 2.851 0.00435 **
## targposAnt:condC 0.3681 0.1917 1.920 0.05491 .
## targposFin:condC 0.2852 0.2238 1.274 0.20269
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 11.4.2 Pseudowords
1.4.2.1 Models for latency
1.4.2.1.1 All targets together
prtn <- lmer ( logRT ~ condC + CprevRT + Corder + (1+Corder|subject) + (cond|name), data3, subset= lex=="Pseudo" & cond!="Neutral", control = lmerControl(optimizer = "bobyqa"))
summary(prtn) # ## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ condC + CprevRT + Corder + (1 + Corder | subject) + (cond |
## name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo" & cond != "Neutral"
##
## REML criterion at convergence: -3426.7
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.6115 -0.6333 -0.1119 0.4894 5.6350
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## name (Intercept) 0.0046512 0.06820
## condMismatch 0.0007588 0.02755 -0.03
## subject (Intercept) 0.0233046 0.15266
## Corder 0.0017856 0.04226 0.15
## Residual 0.0315907 0.17774
## Number of obs: 6868, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.586e+00 1.878e-02 8.706e+01 350.629 < 2e-16 ***
## condC 2.072e-02 4.890e-03 1.341e+02 4.237 4.18e-05 ***
## CprevRT 3.299e-02 2.698e-03 6.715e+03 12.225 < 2e-16 ***
## Corder -1.906e-02 5.374e-03 7.251e+01 -3.547 0.000688 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) condC CprvRT
## condC 0.024
## CprevRT 0.000 -0.017
## Corder 0.126 -0.002 0.0281.4.2.1.2 By target stress position
pintn <- lmer (logRT ~ targpos*condC + CprevRT + Corder + (1+Corder|subject) + (cond|name), data3, subset= lex=="Pseudo" & cond!="Neutral", control = lmerControl(optimizer = "bobyqa"))
spintn <- summary(pintn)
print(spintn,corr=F)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ targpos * condC + CprevRT + Corder + (1 + Corder | subject) +
## (cond | name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo" & cond != "Neutral"
##
## REML criterion at convergence: -3416.7
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.6046 -0.6365 -0.1156 0.4863 5.6439
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## name (Intercept) 0.0040411 0.06357
## condMismatch 0.0007627 0.02762 0.04
## subject (Intercept) 0.0233005 0.15264
## Corder 0.0017891 0.04230 0.15
## Residual 0.0315896 0.17773
## Number of obs: 6868, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.592e+00 2.046e-02 1.170e+02 322.212 < 2e-16 ***
## targposAnt -3.734e-02 1.443e-02 1.366e+02 -2.588 0.010685 *
## targposFin 1.825e-02 1.440e-02 1.359e+02 1.267 0.207215
## condC 2.671e-02 8.467e-03 1.319e+02 3.155 0.001992 **
## CprevRT 3.293e-02 2.698e-03 6.716e+03 12.206 < 2e-16 ***
## Corder -1.915e-02 5.378e-03 7.252e+01 -3.561 0.000657 ***
## targposAnt:condC -1.695e-03 1.200e-02 1.315e+02 -0.141 0.887884
## targposFin:condC -1.610e-02 1.196e-02 1.319e+02 -1.345 0.180793
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 11.4.2.2 Models for accuracy
1.4.2.2.1 All targets together
pern<-glmer(err~condC+(1|subject)+(1|name), data = data3, family = binomial, subset= lex=="Pseudo" & cond!="Neutral", control = glmerControl(optimizer = "bobyqa"))
summary(pern)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: err ~ condC + (1 | subject) + (1 | name)
## Data: data3
## Control: glmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo" & cond != "Neutral"
##
## AIC BIC logLik deviance df.resid
## 1684.0 1711.4 -838.0 1676.0 7100
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.2262 -0.1575 -0.1089 -0.0772 9.8283
##
## Random effects:
## Groups Name Variance Std.Dev.
## name (Intercept) 1.9449 1.3946
## subject (Intercept) 0.3835 0.6192
## Number of obs: 7104, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.47525 0.20027 -22.346 <2e-16 ***
## condC 0.06216 0.15049 0.413 0.68
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## condC -0.0031.4.2.2.2 By target stress position
acpsintn <- glmer (err ~ targpos*condC + (1|subject) + (1|name), data3, family = binomial, subset= lex=="Pseudo" & cond!="Neutral", control = glmerControl(optimizer = "bobyqa"))
acpsintsn <- summary(acpsintn)
print(acpsintsn,corr=F)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: err ~ targpos * condC + (1 | subject) + (1 | name)
## Data: data3
## Control: glmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo" & cond != "Neutral"
##
## AIC BIC logLik deviance df.resid
## 1679.8 1734.7 -831.9 1663.8 7096
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.3282 -0.1572 -0.1063 -0.0749 9.9269
##
## Random effects:
## Groups Name Variance Std.Dev.
## name (Intercept) 1.9641 1.4015
## subject (Intercept) 0.3903 0.6248
## Number of obs: 7104, groups: name, 144; subject, 74
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.6443 0.3060 -15.178 < 2e-16 ***
## targposAnt 0.1751 0.3829 0.457 0.647385
## targposFin 0.2135 0.3804 0.561 0.574671
## condC 0.7777 0.2739 2.840 0.004517 **
## targposAnt:condC -0.8526 0.3719 -2.292 0.021892 *
## targposFin:condC -1.2712 0.3820 -3.328 0.000876 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 11.5 Models for pairwise comparisons of priming effects
Random structures reduced as needed to achieve nonsingular fit. Note: In code here, final stress=“1” ; penultimate=“2” ; antepenultimate=“3”
1.5.1 Words
1.5.1.1 Pen targets
data3$strpos2.12 <- ifelse(data3$strpos=="22",-0.5,ifelse(data3$strpos=="12",0.5,NA))
l2.12 <- lmer ( logRT ~ strpos2.12 + CprevRT + Corder + (strpos2.12|subject) + (0+Corder|subject) + (strpos2.12||name), data3, subset= lex=="Word"&(strpos %in% c("12","22")), control = lmerControl(optimizer = "bobyqa"))
summary(l2.12)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos2.12 + CprevRT + Corder + (strpos2.12 | subject) +
## (0 + Corder | subject) + (strpos2.12 || name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Word" & (strpos %in% c("12", "22"))
##
## REML criterion at convergence: 14
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.6879 -0.6422 -0.0945 0.5191 3.5325
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## subject (Intercept) 0.030049 0.17335
## strpos2.12 0.001289 0.03590 0.24
## subject.1 Corder 0.001440 0.03795
## name (Intercept) 0.017205 0.13117
## name.1 strpos2.12 0.006914 0.08315
## Residual 0.045409 0.21309
## Number of obs: 1522, groups: subject, 74; name, 48
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.581e+00 2.908e-02 1.050e+02 226.273 < 2e-16 ***
## strpos2.12 1.031e-01 2.208e-02 2.320e+01 4.671 0.000104 ***
## CprevRT 3.119e-02 7.252e-03 1.468e+03 4.300 1.82e-05 ***
## Corder -3.919e-03 7.333e-03 6.667e+01 -0.535 0.594769
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st2.12 CprvRT
## strpos2.12 0.213
## CprevRT 0.001 -0.010
## Corder -0.004 0.001 0.074data3$strpos2.32 <- ifelse(data3$strpos=="22",-0.5,ifelse(data3$strpos=="32",0.5,NA))
l2.32 <- lmer ( logRT ~ strpos2.32 + CprevRT + Corder + (1|subject) + (0+Corder|subject) + (strpos2.32|name), data3, subset= lex=="Word"&(strpos %in% c("32","22")))
summary(l2.32)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos2.32 + CprevRT + Corder + (1 | subject) + (0 +
## Corder | subject) + (strpos2.32 | name)
## Data: data3
## Subset: lex == "Word" & (strpos %in% c("32", "22"))
##
## REML criterion at convergence: 15.1
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.7793 -0.6188 -0.1082 0.5037 3.6158
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## subject (Intercept) 0.0290411 0.17041
## subject.1 Corder 0.0006076 0.02465
## name (Intercept) 0.0127523 0.11293
## strpos2.32 0.0043431 0.06590 -0.33
## Residual 0.0470915 0.21701
## Number of obs: 1588, groups: subject, 74; name, 48
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.594e+00 2.671e-02 1.021e+02 246.908 < 2e-16 ***
## strpos2.32 1.301e-01 1.744e-02 2.125e+01 7.459 2.31e-07 ***
## CprevRT 4.844e-02 7.209e-03 1.540e+03 6.719 2.57e-11 ***
## Corder -6.238e-03 6.417e-03 7.421e+01 -0.972 0.334
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st2.32 CprvRT
## strpos2.32 0.020
## CprevRT 0.001 -0.013
## Corder -0.005 -0.001 0.1071.5.1.2 Final targets
data3$strpos1.21 <- ifelse(data3$strpos=="11",-0.5,ifelse(data3$strpos=="21",0.5,NA))
l1.21 <- lmer ( logRT ~ strpos1.21 + CprevRT + Corder + (1|subject) + (0+Corder|subject) + (strpos1.21||name), data3, subset= lex=="Word"&(strpos %in% c("21","11")), control = lmerControl(optimizer = "bobyqa"))
summary(l1.21)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos1.21 + CprevRT + Corder + (1 | subject) + (0 +
## Corder | subject) + (strpos1.21 || name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Word" & (strpos %in% c("21", "11"))
##
## REML criterion at convergence: -345
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.1134 -0.5992 -0.0789 0.4821 5.2891
##
## Random effects:
## Groups Name Variance Std.Dev.
## subject (Intercept) 0.021382 0.14623
## subject.1 Corder 0.001139 0.03374
## name (Intercept) 0.013170 0.11476
## name.1 strpos1.21 0.002765 0.05259
## Residual 0.037812 0.19445
## Number of obs: 1675, groups: subject, 74; name, 48
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.552e+00 2.469e-02 1.069e+02 265.360 < 2e-16 ***
## strpos1.21 9.951e-02 1.555e-02 2.360e+01 6.399 1.39e-06 ***
## CprevRT 5.097e-02 6.082e-03 1.630e+03 8.380 < 2e-16 ***
## Corder 6.207e-03 6.369e-03 7.011e+01 0.975 0.333
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st1.21 CprvRT
## strpos1.21 0.151
## CprevRT 0.001 0.029
## Corder -0.002 -0.017 0.062data3$strpos1.31 <- ifelse(data3$strpos=="11",-0.5,ifelse(data3$strpos=="31",0.5,NA))
l1.31 <- lmer ( logRT ~ strpos1.31 + CprevRT + Corder + (strpos1.31|subject) + (0+Corder|subject) + (strpos1.31|name), data3, subset= lex=="Word"&(strpos %in% c("31","11")))
summary(l1.31)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos1.31 + CprevRT + Corder + (strpos1.31 | subject) +
## (0 + Corder | subject) + (strpos1.31 | name)
## Data: data3
## Subset: lex == "Word" & (strpos %in% c("31", "11"))
##
## REML criterion at convergence: -258.8
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.0972 -0.5869 -0.0999 0.4877 5.4072
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## subject (Intercept) 0.0256296 0.16009
## strpos1.31 0.0008087 0.02844 0.62
## subject.1 Corder 0.0014453 0.03802
## name (Intercept) 0.0104345 0.10215
## strpos1.31 0.0054078 0.07354 -0.29
## Residual 0.0393201 0.19829
## Number of obs: 1662, groups: subject, 74; name, 48
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.547e+00 2.479e-02 9.892e+01 264.130 < 2e-16 ***
## strpos1.31 8.667e-02 1.806e-02 2.418e+01 4.799 6.81e-05 ***
## CprevRT 3.995e-02 6.189e-03 1.601e+03 6.455 1.43e-10 ***
## Corder 6.883e-04 6.808e-03 6.856e+01 0.101 0.92
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st1.31 CprvRT
## strpos1.31 0.129
## CprevRT -0.008 0.002
## Corder 0.000 -0.011 0.0471.5.1.3 Ant targets
data3$strpos3.13 <- ifelse(data3$strpos=="33",-0.5,ifelse(data3$strpos=="13",0.5,NA))
l3.13 <- lmer ( logRT ~ strpos3.13 + CprevRT + Corder + (strpos3.13|subject) + (0+Corder|subject) + (strpos3.13||name), data3, subset= lex=="Word"&(strpos %in% c("13","33")))
summary(l3.13)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos3.13 + CprevRT + Corder + (strpos3.13 | subject) +
## (0 + Corder | subject) + (strpos3.13 || name)
## Data: data3
## Subset: lex == "Word" & (strpos %in% c("13", "33"))
##
## REML criterion at convergence: -116.6
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.7860 -0.6077 -0.1077 0.4956 3.6783
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## subject (Intercept) 0.0289765 0.17022
## strpos3.13 0.0008249 0.02872 0.74
## subject.1 Corder 0.0017356 0.04166
## name (Intercept) 0.0258974 0.16093
## name.1 strpos3.13 0.0001676 0.01295
## Residual 0.0417428 0.20431
## Number of obs: 1556, groups: subject, 74; name, 48
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.644e+00 3.124e-02 9.613e+01 212.691 < 2e-16 ***
## strpos3.13 1.341e-01 1.346e-02 1.899e+01 9.963 5.62e-09 ***
## CprevRT 3.384e-02 6.595e-03 1.502e+03 5.131 3.26e-07 ***
## Corder -1.968e-02 7.356e-03 7.635e+01 -2.675 0.00914 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st3.13 CprvRT
## strpos3.13 0.220
## CprevRT -0.012 -0.079
## Corder -0.003 0.008 0.000data3$strpos3.23 <- ifelse(data3$strpos=="33",-0.5,ifelse(data3$strpos=="23",0.5,NA))
l3.23 <- lmer ( logRT ~ strpos3.23 + CprevRT + Corder + (strpos3.23|subject) + (0+Corder|subject) + (strpos3.23|name), data3, subset= lex=="Word"&(strpos %in% c("23","33")), control = lmerControl(optimizer = "bobyqa"))
summary(l3.23)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos3.23 + CprevRT + Corder + (strpos3.23 | subject) +
## (0 + Corder | subject) + (strpos3.23 | name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Word" & (strpos %in% c("23", "33"))
##
## REML criterion at convergence: 86.6
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.0294 -0.6095 -0.1234 0.5243 3.7455
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## subject (Intercept) 0.030913 0.17582
## strpos3.23 0.003608 0.06007 0.69
## subject.1 Corder 0.001862 0.04315
## name (Intercept) 0.014566 0.12069
## strpos3.23 0.008784 0.09373 -0.70
## Residual 0.047501 0.21795
## Number of obs: 1490, groups: subject, 74; name, 48
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.630e+00 2.792e-02 1.011e+02 237.423 < 2e-16 ***
## strpos3.23 1.097e-01 2.155e-02 3.734e+01 5.091 1.04e-05 ***
## CprevRT 3.175e-02 7.223e-03 1.425e+03 4.395 1.19e-05 ***
## Corder -1.995e-02 7.853e-03 7.229e+01 -2.540 0.0132 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st3.23 CprvRT
## strpos3.23 0.001
## CprevRT 0.010 0.010
## Corder -0.006 0.000 0.0051.5.2 Pseudowords
1.5.2.1 Pen targets
data3$strpos2.12 <- ifelse(data3$strpos=="22",-0.5,ifelse(data3$strpos=="12",0.5,NA))
l2.12 <- lmer ( logRT ~ strpos2.12 + CprevRT + Corder + (strpos2.12|subject) + (0+Corder|subject) + (strpos2.12||name), data3, subset= lex=="Pseudo"&(strpos %in% c("12","22")))
summary(l2.12)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos2.12 + CprevRT + Corder + (strpos2.12 | subject) +
## (0 + Corder | subject) + (strpos2.12 || name)
## Data: data3
## Subset: lex == "Pseudo" & (strpos %in% c("12", "22"))
##
## REML criterion at convergence: -744.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.8402 -0.6012 -0.0924 0.4804 5.4302
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## subject (Intercept) 0.0198750 0.14098
## strpos2.12 0.0001071 0.01035 1.00
## subject.1 Corder 0.0015651 0.03956
## name (Intercept) 0.0048413 0.06958
## name.1 strpos2.12 0.0001413 0.01188
## Residual 0.0305244 0.17471
## Number of obs: 1715, groups: subject, 74; name, 48
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.589e+00 1.992e-02 1.011e+02 330.762 < 2e-16 ***
## strpos2.12 2.278e-02 1.064e-02 2.136e+01 2.140 0.04401 *
## CprevRT 4.710e-02 5.194e-03 1.666e+03 9.068 < 2e-16 ***
## Corder -1.979e-02 6.352e-03 7.302e+01 -3.116 0.00262 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st2.12 CprvRT
## strpos2.12 0.223
## CprevRT -0.004 -0.013
## Corder 0.010 -0.001 0.010data3$strpos2.32 <- ifelse(data3$strpos=="22",-0.5,ifelse(data3$strpos=="32",0.5,NA))
l2.32 <- lmer ( logRT ~ strpos2.32 + CprevRT + Corder + (strpos2.32|subject) + (0+Corder|subject) + (1|name), data3, subset= lex=="Pseudo"&(strpos %in% c("32","22")), control = lmerControl(optimizer = "bobyqa"))
summary(l2.32)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos2.32 + CprevRT + Corder + (strpos2.32 | subject) +
## (0 + Corder | subject) + (1 | name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo" & (strpos %in% c("32", "22"))
##
## REML criterion at convergence: -783.9
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.5317 -0.6136 -0.1056 0.4551 5.6406
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## subject (Intercept) 0.021255 0.14579
## strpos2.32 0.001382 0.03718 0.48
## subject.1 Corder 0.001450 0.03808
## name (Intercept) 0.005791 0.07610
## Residual 0.029661 0.17222
## Number of obs: 1739, groups: subject, 74; name, 48
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.593e+00 2.080e-02 1.042e+02 316.919 < 2e-16 ***
## strpos2.32 3.111e-02 1.087e-02 9.913e+01 2.862 0.00514 **
## CprevRT 4.396e-02 5.099e-03 1.688e+03 8.620 < 2e-16 ***
## Corder -2.464e-02 6.172e-03 6.946e+01 -3.992 0.00016 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st2.32 CprvRT
## strpos2.32 0.259
## CprevRT -0.007 -0.028
## Corder 0.001 -0.032 0.0431.5.2.2 Final targets
data3$strpos1.21 <- ifelse(data3$strpos=="11",-0.5,ifelse(data3$strpos=="21",0.5,NA))
l1.21 <- lmer ( logRT ~ strpos1.21 + CprevRT + Corder + (strpos1.21|subject) + (0+Corder|subject) + (1|name), data3, subset= lex=="Pseudo"&(strpos %in% c("21","11")), control = lmerControl(optimizer = "bobyqa"))
summary(l1.21)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos1.21 + CprevRT + Corder + (strpos1.21 | subject) +
## (0 + Corder | subject) + (1 | name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo" & (strpos %in% c("21", "11"))
##
## REML criterion at convergence: -735.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.4163 -0.5839 -0.1179 0.4760 4.7603
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## subject (Intercept) 0.023102 0.15199
## strpos1.21 0.001493 0.03864 -0.03
## subject.1 Corder 0.001057 0.03250
## name (Intercept) 0.003852 0.06206
## Residual 0.030758 0.17538
## Number of obs: 1729, groups: subject, 74; name, 48
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.605e+00 2.044e-02 9.652e+01 323.132 < 2e-16 ***
## strpos1.21 -2.574e-03 1.105e-02 1.019e+02 -0.233 0.81637
## CprevRT 4.612e-02 5.314e-03 1.681e+03 8.679 < 2e-16 ***
## Corder -1.943e-02 5.811e-03 6.962e+01 -3.344 0.00133 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st1.21 CprvRT
## strpos1.21 0.091
## CprevRT 0.003 0.002
## Corder -0.006 0.008 0.069data3$strpos1.31 <- ifelse(data3$strpos=="11",-0.5,ifelse(data3$strpos=="31",0.5,NA))
l1.31 <- lmer ( logRT ~ strpos1.31 + CprevRT + Corder + (strpos1.31||subject) + (0+Corder|subject) + (strpos1.31||name), data3, subset= lex=="Pseudo"&(strpos %in% c("31","11")), control = lmerControl(optimizer = "bobyqa"))
summary(l1.31)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos1.31 + CprevRT + Corder + (strpos1.31 || subject) +
## (0 + Corder | subject) + (strpos1.31 || name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo" & (strpos %in% c("31", "11"))
##
## REML criterion at convergence: -587.1
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.2785 -0.5831 -0.0970 0.4535 4.3198
##
## Random effects:
## Groups Name Variance Std.Dev.
## subject (Intercept) 0.0229176 0.15139
## subject.1 strpos1.31 0.0019300 0.04393
## subject.2 Corder 0.0017988 0.04241
## name (Intercept) 0.0041912 0.06474
## name.1 strpos1.31 0.0003497 0.01870
## Residual 0.0329026 0.18139
## Number of obs: 1695, groups: subject, 74; name, 48
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.616e+00 2.068e-02 9.729e+01 319.914 < 2e-16 ***
## strpos1.31 2.012e-02 1.246e-02 2.656e+01 1.615 0.11812
## CprevRT 4.707e-02 5.662e-03 1.647e+03 8.313 < 2e-16 ***
## Corder -2.118e-02 6.790e-03 7.734e+01 -3.120 0.00254 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st1.31 CprvRT
## strpos1.31 0.121
## CprevRT -0.001 -0.010
## Corder -0.006 0.007 0.0521.5.2.3 Ant targets
data3$strpos3.13 <- ifelse(data3$strpos=="33",-0.5,ifelse(data3$strpos=="13",0.5,NA))
l3.13 <- lmer ( logRT ~ strpos3.13 + CprevRT + Corder + (strpos3.13||subject) + (0+Corder|subject) + (strpos3.13||name), data3, subset= lex=="Pseudo"&(strpos %in% c("13","33")), control = lmerControl(optimizer = "bobyqa"))
summary(l3.13)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos3.13 + CprevRT + Corder + (strpos3.13 || subject) +
## (0 + Corder | subject) + (strpos3.13 || name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo" & (strpos %in% c("13", "33"))
##
## REML criterion at convergence: -732.1
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.2616 -0.6284 -0.1194 0.4847 5.1838
##
## Random effects:
## Groups Name Variance Std.Dev.
## subject (Intercept) 0.020917 0.14463
## subject.1 strpos3.13 0.000121 0.01100
## subject.2 Corder 0.001988 0.04459
## name (Intercept) 0.003170 0.05630
## name.1 strpos3.13 0.001894 0.04352
## Residual 0.030739 0.17532
## Number of obs: 1741, groups: subject, 74; name, 49
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.564e+00 1.946e-02 9.514e+01 337.395 < 2e-16 ***
## strpos3.13 4.146e-02 1.284e-02 2.648e+01 3.229 0.00331 **
## CprevRT 3.952e-02 5.520e-03 1.698e+03 7.161 1.19e-12 ***
## Corder -1.935e-02 6.741e-03 7.239e+01 -2.870 0.00538 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st3.13 CprvRT
## strpos3.13 0.132
## CprevRT 0.004 -0.002
## Corder 0.002 0.002 0.051data3$strpos3.23 <- ifelse(data3$strpos=="33",-0.5,ifelse(data3$strpos=="23",0.5,NA))
l3.23 <- lmer ( logRT ~ strpos3.23 + CprevRT + Corder + (strpos3.23|subject) + (0+Corder|subject) + (strpos3.23|name), data3, subset= lex=="Pseudo"&(strpos %in% c("23","33")), control = lmerControl(optimizer = "bobyqa"))
summary(l3.23)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: logRT ~ strpos3.23 + CprevRT + Corder + (strpos3.23 | subject) +
## (0 + Corder | subject) + (strpos3.23 | name)
## Data: data3
## Control: lmerControl(optimizer = "bobyqa")
## Subset: lex == "Pseudo" & (strpos %in% c("23", "33"))
##
## REML criterion at convergence: -692
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.6007 -0.6403 -0.1066 0.4637 4.8517
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## subject (Intercept) 0.022535 0.15012
## strpos3.23 0.002750 0.05244 0.55
## subject.1 Corder 0.001851 0.04302
## name (Intercept) 0.001932 0.04396
## strpos3.23 0.001635 0.04043 -0.72
## Residual 0.030931 0.17587
## Number of obs: 1687, groups: subject, 74; name, 48
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.547e+00 1.923e-02 8.317e+01 340.504 < 2e-16 ***
## strpos3.23 8.777e-03 1.304e-02 3.446e+01 0.673 0.50548
## CprevRT 3.704e-02 5.646e-03 1.627e+03 6.560 7.2e-11 ***
## Corder -1.915e-02 6.722e-03 6.683e+01 -2.849 0.00582 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) st3.23 CprvRT
## strpos3.23 0.210
## CprevRT 0.001 -0.014
## Corder 0.001 -0.003 0.0371.6 Bayes factor for the interaction between stress position and stress priming
To quantify support for the null (i.e., no difference in stress priming among stress patterns) we compute a Bayes factor comparing two models, one with the interaction and one without.
data33 <- na.omit(data3[data3$lex=="Word" & data3$cond!="Neutral", # lmBF cannot deal with missing values
c("targpos","condC","CprevRT","Corder","subject","name","logRT")])
data33$sID <- data33$subject # fixing nul in string
lb1 <- lmBF( logRT ~ targpos*condC + CprevRT + Corder + sID + name, data33, whichRandom=c("sID","name"))
lb0 <- lmBF( logRT ~ targpos+condC + CprevRT + Corder + sID + name, data33, whichRandom=c("sID","name"))
lb1/lb0## Bayes factor analysis
## --------------
## [1] targpos * condC + CprevRT + Corder + sID + name : 0.09654998 ±2.47%
##
## Against denominator:
## logRT ~ targpos + condC + CprevRT + Corder + sID + name
## ---
## Bayes factor type: BFlinearModel, JZSThe value 0.1 indicates moderate to strong support for the null.
1.7 Session information
sessionInfo()## R version 3.6.2 (2019-12-12)
## Platform: x86_64-w64-mingw32/x64 (64-bit)
## Running under: Windows 10 x64 (build 18363)
##
## Matrix products: default
##
## locale:
## [1] LC_COLLATE=English_United States.1252
## [2] LC_CTYPE=English_United States.1252
## [3] LC_MONETARY=English_United States.1252
## [4] LC_NUMERIC=C
## [5] LC_TIME=English_United States.1252
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] BayesFactor_0.9.12-4.2 coda_0.19-3 dfoptim_2018.2-1
## [4] gtools_3.8.1 effects_4.1-4 carData_3.0-3
## [7] lmerTest_3.1-1 lattice_0.20-40 lme4_1.1-21
## [10] Matrix_1.2-18
##
## loaded via a namespace (and not attached):
## [1] tidyselect_0.2.5 xfun_0.12 purrr_0.3.3
## [4] pbapply_1.4-2 mitools_2.4 splines_3.6.2
## [7] colorspace_1.4-1 htmltools_0.4.0 yaml_2.2.1
## [10] survival_3.1-8 rlang_0.4.2 nloptr_1.2.1
## [13] pillar_1.4.3 glue_1.3.1 DBI_1.1.0
## [16] lifecycle_0.1.0 stringr_1.4.0 MatrixModels_0.4-1
## [19] munsell_0.5.0 gtable_0.3.0 mvtnorm_1.0-11
## [22] evaluate_0.14 knitr_1.28 parallel_3.6.2
## [25] Rcpp_1.0.3 scales_1.1.0 ggplot2_3.2.1
## [28] digest_0.6.25 stringi_1.4.6 dplyr_0.8.3
## [31] survey_3.36 numDeriv_2016.8-1.1 grid_3.6.2
## [34] tools_3.6.2 magrittr_1.5 lazyeval_0.2.2
## [37] tibble_2.1.3 crayon_1.3.4 pkgconfig_2.0.3
## [40] MASS_7.3-51.5 estimability_1.3 assertthat_0.2.1
## [43] minqa_1.2.4 rmarkdown_2.1 R6_2.4.1
## [46] boot_1.3-24 nnet_7.3-13 nlme_3.1-144
## [49] compiler_3.6.2