# LuCiD WP8 Study 1: Longitudinal study # GLMER-analysis of predictors of False Belief understanding at Time 2 (stepwise backwards selection) library(lme4) lucid1<-read.csv2(file.choose()) # Choose "LuCiD_WP8_Study1_T2_EFB.csv" # First full model including all experimental and control variables: luc1a.glmer <-glmer(EFB2 ~ EFB1+ToM+Age2+Gender+Sibling+STM+WM+DCCS+ICmot+ICverb+Vocab+Gram+Complements+MV1+MV3+Modals+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 204.8 269.8 -80.4 160.8 120 # # Scaled residuals: # Min 1Q Median 3Q Max # -3.4304 -0.6870 0.1386 0.7313 2.0553 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 2.194e-06 0.001481 # Item (Intercept) 5.453e-01 0.738439 # Experimenter (Intercept) 2.425e-06 0.001557 # Set (Intercept) 2.754e-06 0.001660 # Order (Intercept) 1.628e-06 0.001276 # Number of obs: 142, groups: ID, 30; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) 2.79672 4.55488 0.614 0.5392 # EFB1 1.12285 0.46985 2.390 0.0169 * # ToM 0.39815 0.52341 0.761 0.4468 # Age2 -0.13883 0.12330 -1.126 0.2602 # GenderM 0.70206 0.59886 1.172 0.2411 # Siblingyounger -0.33318 0.58471 -0.570 0.5688 # STM 0.15265 0.10793 1.414 0.1573 # WM 0.24791 0.26595 0.932 0.3512 # DCCS -0.02373 0.11035 -0.215 0.8297 # ICmot 0.17903 0.29270 0.612 0.5408 # ICverb -0.04181 0.07502 -0.557 0.5774 # Vocab -0.06340 0.03039 -2.086 0.0370 * # Gram 0.10443 0.06788 1.538 0.1239 # Complements 0.15814 0.06273 2.521 0.0117 * # MV1 -0.94284 0.62089 -1.519 0.1289 # MV3 0.38399 0.36154 1.062 0.2882 # Modals 0.05365 0.56507 0.095 0.9244 # Modals discarded: luc1b.glmer <-glmer(EFB2 ~ EFB1+ToM+Age2+Gender+Sibling+STM+WM+DCCS+ICmot+ICverb+Vocab+Gram+Complements+MV1+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 202.8 264.9 -80.4 160.8 121 # # Scaled residuals: # Min 1Q Median 3Q Max # -3.4109 -0.6937 0.1388 0.7311 2.0719 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 3.289e-05 0.005735 # Item (Intercept) 5.482e-01 0.740406 # Experimenter (Intercept) 7.622e-05 0.008731 # Set (Intercept) 2.342e-05 0.004839 # Order (Intercept) 2.038e-05 0.004515 # Number of obs: 142, groups: ID, 30; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) 2.81459 4.53208 0.621 0.5346 # EFB1 1.11570 0.46347 2.407 0.0161 * # ToM 0.36961 0.46212 0.800 0.4238 # Age2 -0.14034 0.12080 -1.162 0.2453 # GenderM 0.68809 0.59054 1.165 0.2439 # Siblingyounger -0.32119 0.56534 -0.568 0.5699 # STM 0.14867 0.10193 1.459 0.1447 # WM 0.25977 0.23422 1.109 0.2674 # DCCS -0.02695 0.10173 -0.265 0.7911 # ICmot 0.19236 0.24411 0.788 0.4307 # ICverb -0.04440 0.06751 -0.658 0.5107 # Vocab -0.06152 0.02527 -2.434 0.0149 * # Gram 0.10409 0.06766 1.538 0.1240 # Complements 0.15780 0.06230 2.533 0.0113 * # MV1 -0.90214 0.50420 -1.789 0.0736 . # MV3 0.37211 0.33971 1.095 0.2734 # No effect of discarding Modals: anova(luc1a.glmer, luc1b.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1b.glmer 21 202.83 264.90 -80.413 160.83 # luc1a.glmer 22 204.82 269.85 -80.409 160.82 0.0095 1 0.9224 # DCCS (Cognitive Flexibility) discarded: luc1c.glmer <-glmer(EFB2 ~ EFB1+ToM+Age2+Gender+Sibling+STM+WM+ICmot+ICverb+Vocab+Gram+Complements+MV1+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 214.1 274.0 -87.0 174.1 128 # # Scaled residuals: # Min 1Q Median 3Q Max # -2.9689 -0.7095 0.1696 0.7221 2.3176 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 3.031e-07 0.0005505 # Item (Intercept) 4.061e-01 0.6372380 # Experimenter (Intercept) 3.099e-07 0.0005567 # Set (Intercept) 3.201e-04 0.0178915 # Order (Intercept) 1.167e-05 0.0034160 # Number of obs: 148, groups: ID, 32; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -1.57697 4.17439 -0.378 0.7056 # EFB1 0.94092 0.42487 2.215 0.0268 * # ToM 0.22162 0.43204 0.513 0.6080 # Age2 -0.02006 0.10441 -0.192 0.8476 # GenderM 0.36125 0.51721 0.698 0.4849 # Siblingyounger -0.09533 0.47773 -0.200 0.8418 # STM 0.06041 0.09480 0.637 0.5240 # WM 0.12657 0.21425 0.591 0.5547 # ICmot 0.25681 0.22554 1.139 0.2549 # ICverb -0.03473 0.05803 -0.598 0.5495 # Vocab -0.04314 0.02256 -1.912 0.0558 . # Gram 0.09422 0.06452 1.460 0.1442 # Complements 0.09840 0.04979 1.976 0.0482 * # MV1 -0.33380 0.43158 -0.773 0.4393 # MV3 0.30277 0.33318 0.909 0.3635 anova(luc1b.glmer, luc1c.glmer) # Error in anova.merMod(luc1b.glmer, luc1c.glmer) : # models were not all fitted to the same size of dataset # [Error in comparison due to three children having NA's on DCCS] # Fitting the models to the same data sets without NA's lucid1.na.omit = data.frame(lucid1$EFB2, lucid1$EFB1,lucid1$ToM, lucid1$Age2,lucid1$Gender,lucid1$Sibling,lucid1$STM,lucid1$WM,lucid1$DCCS,lucid1$ICmot,lucid1$ICverb, lucid1$Vocab, lucid1$Gram, lucid1$Complements, lucid1$MV1, lucid1$MV3, lucid1$ID, lucid1$Item, lucid1$Experimenter, lucid1$Set, lucid1$Order) lucid1.na.omit = droplevels (data.frame(na.omit(lucid1.na.omit))) luc1b.na.omit.glmer <-glmer(lucid1.EFB2 ~ lucid1.EFB1+lucid1.ToM+lucid1.Age2+lucid1.Gender+lucid1.Sibling+lucid1.STM+lucid1.WM+lucid1.DCCS+lucid1.ICmot+lucid1.ICverb+lucid1.Vocab+lucid1.Gram+lucid1.Complements+lucid1.MV1+lucid1.MV3+(1|lucid1.ID)+(1|lucid1.Item)+(1|lucid1.Experimenter)+(1|lucid1.Set)+(1|lucid1.Order), family = "binomial", data = lucid1.na.omit) luc1c.na.omit.glmer <-glmer(lucid1.EFB2 ~ lucid1.EFB1+lucid1.ToM+lucid1.Age2+lucid1.Gender+lucid1.Sibling+lucid1.STM+lucid1.WM+lucid1.ICmot+lucid1.ICverb+lucid1.Vocab+lucid1.Gram+lucid1.Complements+lucid1.MV1+lucid1.MV3+(1|lucid1.ID)+(1|lucid1.Item)+(1|lucid1.Experimenter)+(1|lucid1.Set)+(1|lucid1.Order), family = "binomial", data = lucid1.na.omit) # Model comparison with models fitted to the same data set, no effect of discarding DCCS (Cognitive Flexibility): anova(luc1b.na.omit.glmer, luc1c.na.omit.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1c.na.omit.glmer 20 200.90 260.02 -80.450 160.90 # luc1b.na.omit.glmer 21 202.83 264.90 -80.413 160.83 0.0736 1 0.7862 # Age at Time 2 discarded: luc1d.glmer <-glmer(EFB2 ~ EFB1+ToM+Gender+Sibling+STM+WM+ICmot+ICverb+Vocab+Gram+Complements+MV1+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 212.1 269.0 -87.0 174.1 129 # # Scaled residuals: # Min 1Q Median 3Q Max # -2.9669 -0.7132 0.1680 0.7251 2.2242 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 1.266e-07 0.0003558 # Item (Intercept) 4.101e-01 0.6404182 # Experimenter (Intercept) 8.886e-06 0.0029810 # Set (Intercept) 9.625e-03 0.0981090 # Order (Intercept) 1.457e-05 0.0038168 # Number of obs: 148, groups: ID, 32; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -2.38384 1.72205 -1.384 0.1663 # EFB1 0.93695 0.43082 2.175 0.0296 * # ToM 0.24663 0.46858 0.526 0.5987 # GenderM 0.34604 0.53045 0.652 0.5142 # Siblingyounger -0.10625 0.47655 -0.223 0.8236 # STM 0.06148 0.10273 0.598 0.5496 # WM 0.11310 0.20391 0.555 0.5791 # ICmot 0.25826 0.22556 1.145 0.2522 # ICverb -0.03660 0.06067 -0.603 0.5464 # Vocab -0.04350 0.02282 -1.906 0.0566 . # Gram 0.09537 0.07759 1.229 0.2190 # Complements 0.09471 0.04776 1.983 0.0474 * # MV1 -0.34249 0.43834 -0.781 0.4346 # MV3 0.31535 0.33625 0.938 0.3483 # No effect of discarding Age at Time 2: # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1d.glmer 19 212.09 269.04 -87.044 174.09 # luc1c.glmer 20 214.05 274.00 -87.027 174.05 0.0347 1 0.8523 # Sibling status discarded: luc1e.glmer <-glmer(EFB2 ~ EFB1+ToM+Gender+STM+WM+ICmot+ICverb+Vocab+Gram+Complements+MV1+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 210.1 264.1 -87.1 174.1 130 # # Scaled residuals: # Min 1Q Median 3Q Max # -3.1465 -0.7005 0.1613 0.7256 2.1691 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 8.232e-06 0.002869 # Item (Intercept) 4.087e-01 0.639313 # Experimenter (Intercept) 7.769e-06 0.002787 # Set (Intercept) 6.001e-03 0.077469 # Order (Intercept) 1.178e-05 0.003433 # Number of obs: 148, groups: ID, 32; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -2.39993 1.73827 -1.381 0.1674 # EFB1 0.93336 0.43088 2.166 0.0303 * # ToM 0.27461 0.45412 0.605 0.5454 # GenderM 0.33576 0.53219 0.631 0.5281 # STM 0.05540 0.09930 0.558 0.5770 # WM 0.11218 0.20434 0.549 0.5830 # ICmot 0.27235 0.21481 1.268 0.2049 # ICverb -0.04074 0.05772 -0.706 0.4802 # Vocab -0.04422 0.02263 -1.954 0.0507 . # Gram 0.09259 0.07709 1.201 0.2297 # Complements 0.09787 0.04608 2.124 0.0337 * # MV1 -0.33400 0.43832 -0.762 0.4461 # MV3 0.31001 0.33592 0.923 0.3561 # No effect of discarding Sibling status: anova(luc1d.glmer, luc1e.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1e.glmer 18 210.14 264.09 -87.070 174.14 # luc1d.glmer 19 212.09 269.04 -87.044 174.09 0.0508 1 0.8216 # Working Memory (Missing Scan Task) discarded: luc1f.glmer <-glmer(EFB2 ~ EFB1+ToM+Gender+STM+ICmot+ICverb+Vocab+Gram+Complements+MV1+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 208.5 259.4 -87.2 174.5 131 # # Scaled residuals: # Min 1Q Median 3Q Max # -3.2338 -0.7256 0.1649 0.7301 2.1154 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 3.841e-07 0.0006198 # Item (Intercept) 3.986e-01 0.6313284 # Experimenter (Intercept) 2.422e-06 0.0015564 # Set (Intercept) 3.019e-07 0.0005494 # Order (Intercept) 6.751e-06 0.0025982 # Number of obs: 148, groups: ID, 32; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -2.01867 1.19094 -1.695 0.0901 . # EFB1 0.87779 0.40914 2.145 0.0319 * # ToM 0.34142 0.36225 0.943 0.3459 # GenderM 0.34480 0.49294 0.699 0.4843 # STM 0.05846 0.08881 0.658 0.5104 # ICmot 0.26548 0.21405 1.240 0.2149 # ICverb -0.03613 0.05283 -0.684 0.4940 # Vocab -0.04365 0.02222 -1.965 0.0495 * # Gram 0.07703 0.05512 1.398 0.1622 # Complements 0.09731 0.04520 2.153 0.0313 * # MV1 -0.37305 0.42351 -0.881 0.3784 # MV3 0.37539 0.30473 1.232 0.2180 # No effect of discarding Working Memory: anova(luc1e.glmer, luc1f.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1f.glmer 17 208.47 259.42 -87.234 174.47 # luc1e.glmer 18 210.14 264.09 -87.070 174.14 0.3294 1 0.566 # Short-Term Memory (Digit Span Forwards) discarded: luc1g.glmer <-glmer(EFB2 ~ EFB1+ToM+Gender+ICmot+ICverb+Vocab+Gram+Complements+MV1+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 206.9 254.9 -87.5 174.9 132 # # Scaled residuals: # Min 1Q Median 3Q Max # -3.1338 -0.6958 0.1667 0.7635 2.0217 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 4.226e-06 0.002056 # Item (Intercept) 4.058e-01 0.637054 # Experimenter (Intercept) 2.278e-05 0.004773 # Set (Intercept) 6.475e-06 0.002545 # Order (Intercept) 1.074e-05 0.003278 # Number of obs: 148, groups: ID, 32; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -2.12158 1.17787 -1.801 0.0717 . # EFB1 0.90738 0.40591 2.235 0.0254 * # ToM 0.40296 0.35012 1.151 0.2498 # GenderM 0.35382 0.49002 0.722 0.4703 # ICmot 0.33699 0.18457 1.826 0.0679 . # ICverb -0.03286 0.05251 -0.626 0.5315 # Vocab -0.04206 0.02197 -1.915 0.0555 . # Gram 0.07802 0.05524 1.413 0.1578 # Complements 0.09689 0.04486 2.160 0.0308 * # MV1 -0.24781 0.37683 -0.658 0.5108 # MV3 0.38285 0.30256 1.265 0.2057 # No effect of discarding Short-Term Memory: anova(luc1f.glmer, luc1g.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1g.glmer 16 206.91 254.86 -87.452 174.91 # luc1f.glmer 17 208.47 259.42 -87.234 174.47 0.4364 1 0.5089 # Verbal Inhibitory Control (Black/White task) discarded: luc1h.glmer <-glmer(EFB2 ~ EFB1+ToM+Gender+ICmot+Vocab+Gram+Complements+MV1+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 214.3 259.8 -92.1 184.3 138 # # Scaled residuals: # Min 1Q Median 3Q Max # -2.1913 -0.7230 0.2351 0.7329 1.9195 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 3.125e-06 0.001768 # Item (Intercept) 3.969e-01 0.629984 # Experimenter (Intercept) 5.115e-06 0.002262 # Set (Intercept) 3.714e-06 0.001927 # Order (Intercept) 3.357e-06 0.001832 # Number of obs: 153, groups: ID, 33; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -1.57910 1.10957 -1.423 0.1547 # EFB1 0.86197 0.39754 2.168 0.0301 * # ToM 0.27560 0.33977 0.811 0.4173 # GenderM 0.20902 0.40372 0.518 0.6046 # ICmot 0.20943 0.15975 1.311 0.1899 # Vocab -0.03405 0.02083 -1.634 0.1022 # Gram 0.06161 0.05291 1.164 0.2443 # Complements 0.06819 0.03652 1.867 0.0619 . # MV1 -0.20007 0.34389 -0.582 0.5607 # MV3 0.54655 0.27345 1.999 0.0456 * # Model comparison anova(luc1g.glmer, luc1h.glmer) # Error in anova.merMod(luc1g.glmer, luc1h.glmer) : # models were not all fitted to the same size of dataset # [Error in comparison due to one child having NA on ICverb] # Fitting the models to the same data sets without NA's lucid1.na.omit = data.frame(lucid1$EFB2, lucid1$EFB1,lucid1$ToM, lucid1$Gender,lucid1$ICmot,lucid1$ICverb, lucid1$Vocab, lucid1$Gram, lucid1$Complements, lucid1$MV1, lucid1$MV3, lucid1$ID, lucid1$Item, lucid1$Experimenter, lucid1$Set, lucid1$Order) lucid1.na.omit = droplevels (data.frame(na.omit(lucid1.na.omit))) luc1g.na.omit.glmer <-glmer(lucid1.EFB2 ~ lucid1.EFB1+lucid1.ToM+lucid1.Gender+lucid1.ICmot+lucid1.ICverb+lucid1.Vocab+lucid1.Gram+lucid1.Complements+lucid1.MV1+lucid1.MV3+(1|lucid1.ID)+(1|lucid1.Item)+(1|lucid1.Experimenter)+(1|lucid1.Set)+(1|lucid1.Order), family = "binomial", data = lucid1.na.omit) luc1h.na.omit.glmer <-glmer(lucid1.EFB2 ~ lucid1.EFB1+lucid1.ToM+lucid1.Gender+lucid1.ICmot+lucid1.Vocab+lucid1.Gram+lucid1.Complements+lucid1.MV1+lucid1.MV3+(1|lucid1.ID)+(1|lucid1.Item)+(1|lucid1.Experimenter)+(1|lucid1.Set)+(1|lucid1.Order), family = "binomial", data = lucid1.na.omit) # Model comparison with models fitted to the same data set, no effect of discarding Verbal Inhibitory Control (Black/White task): anova(luc1g.na.omit.glmer, luc1h.na.omit.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1h.na.omit.glmer 15 205.31 250.26 -87.653 175.31 # luc1g.na.omit.glmer 16 206.91 254.86 -87.452 174.91 0.401 1 0.5266 # Gender discarded: luc1i.glmer <-glmer(EFB2 ~ EFB1+ToM+ICmot+Vocab+Gram+Complements+MV1+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 212.6 255.0 -92.3 184.6 139 # # Scaled residuals: # Min 1Q Median 3Q Max # -2.2791 -0.7136 0.2269 0.7304 1.8479 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 5.935e-06 0.002436 # Item (Intercept) 3.906e-01 0.624954 # Experimenter (Intercept) 1.638e-05 0.004047 # Set (Intercept) 1.059e-05 0.003254 # Order (Intercept) 1.196e-06 0.001093 # Number of obs: 153, groups: ID, 33; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -1.37460 1.02670 -1.339 0.1806 # EFB1 0.89638 0.39284 2.282 0.0225 * # ToM 0.23434 0.32887 0.713 0.4761 # ICmot 0.18686 0.15379 1.215 0.2243 # Vocab -0.03403 0.02074 -1.641 0.1008 # Gram 0.06613 0.05224 1.266 0.2055 # Complements 0.06671 0.03639 1.833 0.0668 . # MV1 -0.21357 0.34224 -0.624 0.5326 # MV3 0.55768 0.27332 2.040 0.0413 * # No effect of discarding Gender: anova(luc1h.glmer, luc1i.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1i.glmer 14 212.56 254.99 -92.281 184.56 # luc1h.glmer 15 214.30 259.75 -92.149 184.30 0.265 1 0.6067 # Mental Verbs with First-person complements discarded: luc1j.glmer <-glmer(EFB2 ~ EFB1+ToM+ICmot+Vocab+Gram+Complements+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 211.0 250.4 -92.5 185.0 140 # # Scaled residuals: # Min 1Q Median 3Q Max # -2.0468 -0.7004 0.2560 0.7579 1.8097 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 3.725e-06 0.001930 # Item (Intercept) 3.832e-01 0.618993 # Experimenter (Intercept) 1.450e-05 0.003808 # Set (Intercept) 1.021e-05 0.003196 # Order (Intercept) 2.372e-06 0.001540 # Number of obs: 153, groups: ID, 33; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -1.52145 0.99702 -1.526 0.1270 # EFB1 0.86937 0.38943 2.232 0.0256 * # ToM 0.19219 0.31901 0.602 0.5469 # ICmot 0.19051 0.15335 1.242 0.2141 # Vocab -0.02836 0.01856 -1.528 0.1266 # Gram 0.05901 0.05089 1.160 0.2462 # Complements 0.06002 0.03452 1.739 0.0820 . # MV3 0.55013 0.27227 2.021 0.0433 * # No effect of discarding Mental Verbs with First-person complements: anova(luc1i.glmer, luc1j.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1j.glmer 13 210.96 250.35 -92.479 184.96 # luc1i.glmer 14 212.56 254.99 -92.281 184.56 0.3962 1 0.5291 # Theory of Mind Precursors (Diverse Desires, Diverse Beliefs) discarded: luc1k.glmer <-glmer(EFB2 ~ EFB1+ICmot+Vocab+Gram+Complements+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 209.3 245.7 -92.7 185.3 141 # # Scaled residuals: # Min 1Q Median 3Q Max # -2.0645 -0.7315 0.2689 0.7788 1.8497 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 1.381e-06 1.175e-03 # Item (Intercept) 3.757e-01 6.129e-01 # Experimenter (Intercept) 2.583e-08 1.607e-04 # Set (Intercept) 2.488e-09 4.988e-05 # Order (Intercept) 6.032e-09 7.767e-05 # Number of obs: 153, groups: ID, 33; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -1.17662 0.80678 -1.458 0.1447 # EFB1 0.86287 0.38927 2.217 0.0266 * # ICmot 0.19223 0.15298 1.257 0.2089 # Vocab -0.02880 0.01853 -1.555 0.1200 # Gram 0.06018 0.05084 1.184 0.2365 # Complements 0.05830 0.03418 1.706 0.0881 . # MV3 0.49086 0.25114 1.955 0.0506 . # No effect of discarding Theory of Mind Precursors: anova(luc1j.glmer, luc1k.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1k.glmer 12 209.32 245.69 -92.660 185.32 # luc1j.glmer 13 210.96 250.35 -92.479 184.96 0.3626 1 0.547 # Receptive Grammar (Sentence Structure from CELF) discarded: luc1l.glmer <-glmer(EFB2 ~ EFB1+ICmot+Vocab+Complements+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 208.7 242.1 -93.4 186.7 142 # # Scaled residuals: # Min 1Q Median 3Q Max # -2.0601 -0.7597 0.3016 0.7493 2.1244 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 3.381e-08 0.0001839 # Item (Intercept) 3.676e-01 0.6063403 # Experimenter (Intercept) 9.945e-07 0.0009973 # Set (Intercept) 1.551e-07 0.0003938 # Order (Intercept) 7.072e-07 0.0008409 # Number of obs: 153, groups: ID, 33; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -0.96780 0.79061 -1.224 0.2209 # EFB1 0.80567 0.38177 2.110 0.0348 * # ICmot 0.26682 0.14122 1.889 0.0588 . # Vocab -0.02768 0.01842 -1.502 0.1330 # Complements 0.06669 0.03325 2.006 0.0449 * # MV3 0.47187 0.24782 1.904 0.0569 . # No effect of discarding Receptive Grammar anova(luc1k.glmer, luc1l.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1l.glmer 11 208.74 242.08 -93.371 186.74 # luc1k.glmer 12 209.32 245.69 -92.660 185.32 1.4202 1 0.2334 # Receptive Vocabulary (BPVS) discarded: luc1m.glmer <-glmer(EFB2 ~ EFB1+ICmot+Complements+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 209.1 239.4 -94.5 189.1 143 # # Scaled residuals: # Min 1Q Median 3Q Max # -1.7009 -0.8031 0.2868 0.8003 2.0778 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 1.069e-05 0.0032690 # Item (Intercept) 3.641e-01 0.6034213 # Experimenter (Intercept) 5.672e-07 0.0007531 # Set (Intercept) 4.428e-06 0.0021042 # Order (Intercept) 2.790e-06 0.0016703 # Number of obs: 153, groups: ID, 33; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -1.67816 0.64501 -2.602 0.00928 ** # EFB1 0.80068 0.37891 2.113 0.03459 * # ICmot 0.17523 0.12828 1.366 0.17193 # Complements 0.04432 0.02878 1.540 0.12358 # MV3 0.54286 0.23906 2.271 0.02316 * # No effect of discarding Receptive Vocabulary: anova(luc1l.glmer, luc1m.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1m.glmer 10 209.06 239.37 -94.532 189.06 # luc1l.glmer 11 208.74 242.08 -93.371 186.74 2.3228 1 0.1275 # Motoric Inhibitory Control (Sheep/Crocodile task) discarded: luc1n.glmer <-glmer(EFB2 ~ EFB1+Complements+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 272.1 301.8 -127.0 254.1 191 # # Scaled residuals: # Min 1Q Median 3Q Max # -1.8074 -0.8156 -0.5089 0.8857 1.9399 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 1.559e-07 3.949e-04 # Item (Intercept) 2.367e-01 4.865e-01 # Experimenter (Intercept) 1.263e-10 1.124e-05 # Set (Intercept) 1.836e-09 4.285e-05 # Order (Intercept) 1.122e-10 1.059e-05 # Number of obs: 200, groups: ID, 44; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -1.00058 0.37085 -2.698 0.00697 ** # EFB1 0.71189 0.32167 2.213 0.02689 * # Complements 0.05160 0.02407 2.144 0.03206 * # MV3 0.45392 0.21964 2.067 0.03877 * # Model comparison anova(luc1m.glmer, luc1n.glmer) # Error in anova.merMod(luc1m.glmer, luc1n.glmer) : # models were not all fitted to the same size of dataset # [Error in comparison due to 11 children having NA's on ICmot] # Fitting the models to the same data sets without NA's lucid1.na.omit = data.frame(lucid1$EFB2, lucid1$EFB1,lucid1$Complements, lucid1$MV3, lucid1$ICmot, lucid1$ID, lucid1$Item, lucid1$Experimenter, lucid1$Set, lucid1$Order) lucid1.na.omit = droplevels (data.frame(na.omit(lucid1.na.omit))) luc1m.na.omit.glmer <- glmer(lucid1.EFB2 ~ lucid1.EFB1+lucid1.Complements+lucid1.MV3+lucid1.ICmot+(1|lucid1.ID)+(1|lucid1.Item)+(1|lucid1.Experimenter)+(1|lucid1.Set)+(1|lucid1.Order), family = "binomial", data = lucid1.na.omit) luc1n.na.omit.glmer <-glmer(lucid1.EFB2 ~ lucid1.EFB1+lucid1.Complements+lucid1.MV3+(1|lucid1.ID)+(1|lucid1.Item)+(1|lucid1.Experimenter)+(1|lucid1.Set)+(1|lucid1.Order), family = "binomial", data = lucid1.na.omit) # Model comparison with models fitted to the same data set, no effect of discarding Motoric Inhibitory Control (Sheep/Crocodile task): anova(luc1m.na.omit.glmer, luc1n.na.omit.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1n.na.omit.glmer 9 209.02 236.30 -95.511 191.02 # luc1m.na.omit.glmer 10 209.06 239.37 -94.532 189.06 1.959 1 0.1616 # Mental Verbs with Third-person complements discarded: luc1o.glmer <-glmer(EFB2 ~ EFB1+Complements+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1) # AIC BIC logLik deviance df.resid # 274.4 300.8 -129.2 258.4 192 # # Scaled residuals: # Min 1Q Median 3Q Max # -1.8695 -0.8222 -0.5436 0.8844 1.8158 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 0.000e+00 0.000e+00 # Item (Intercept) 2.321e-01 4.818e-01 # Experimenter (Intercept) 1.768e-02 1.330e-01 # Set (Intercept) 1.128e-10 1.062e-05 # Order (Intercept) 2.346e-10 1.532e-05 # Number of obs: 200, groups: ID, 44; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -0.82695 0.37596 -2.200 0.0278 * # EFB1 0.72392 0.32381 2.236 0.0254 * # Complements 0.05224 0.02458 2.126 0.0335 * # Model comparison anova(luc1n.glmer, luc1o.glmer) # Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) # luc1o.glmer 8 274.41 300.79 -129.20 258.41 # luc1n.glmer 9 272.09 301.77 -127.04 254.09 4.3167 1 0.03774 * # Discarding Mental Verbs with Third-person complements would make the model significantly worse. # Therefore luc1n.glmer is the final model: # AIC BIC logLik deviance df.resid # 272.1 301.8 -127.0 254.1 191 # # Scaled residuals: # Min 1Q Median 3Q Max # -1.8074 -0.8156 -0.5089 0.8857 1.9399 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 1.559e-07 3.949e-04 # Item (Intercept) 2.367e-01 4.865e-01 # Experimenter (Intercept) 1.263e-10 1.124e-05 # Set (Intercept) 1.836e-09 4.285e-05 # Order (Intercept) 1.122e-10 1.059e-05 # Number of obs: 200, groups: ID, 44; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -1.00058 0.37085 -2.698 0.00697 ** # EFB1 0.71189 0.32167 2.213 0.02689 * # Complements 0.05160 0.02407 2.144 0.03206 * # MV3 0.45392 0.21964 2.067 0.03877 * # FOLLOW-UP ANALYSES # Check for outliers (large standardised residuals: >2.5) # No outliers: same model. luc1n.glmerB <-glmer(EFB2 ~ EFB1+Complements+MV3+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1[abs(scale(resid(luc1n.glmer)))< 2.5,]) summary(luc1n.glmerB) # AIC BIC logLik deviance df.resid # 272.1 301.8 -127.0 254.1 191 # # Scaled residuals: # Min 1Q Median 3Q Max # -1.8074 -0.8156 -0.5089 0.8857 1.9399 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 1.559e-07 3.949e-04 # Item (Intercept) 2.367e-01 4.865e-01 # Experimenter (Intercept) 1.263e-10 1.124e-05 # Set (Intercept) 1.836e-09 4.285e-05 # Order (Intercept) 1.122e-10 1.059e-05 # Number of obs: 200, groups: ID, 44; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -1.00058 0.37085 -2.698 0.00697 ** # EFB1 0.71189 0.32167 2.213 0.02689 * # Complements 0.05160 0.02407 2.144 0.03206 * # MV3 0.45392 0.21964 2.067 0.03877 * # Marginal effects plot library(ggeffects) plot(ggpredict(luc1n.glmer, terms = c("Complements", "MV3", "EFB1"))) # Post hoc power # Make sure to prune the data set for rows containing NA's before conducting the simulations, otherwise every simulation will fail, and you'll see observed powers of zero all way. library(simr) powerSim(luc1n.glmer, fixed("EFB1", "z"), nsim=1000) powerSim(luc1n.glmer, fixed("Complements", "z"), nsim=1000) powerSim(luc1n.glmer, fixed("MV3", "z"), nsim=1000) # No interaction between effect of Mental Verbs with Third-person complements and use of the mental verb Think in the FB test question: summary(glmer(EFB2 ~ EFB1+Complements+MV3*Think+(1|ID)+(1|Item)+(1|Experimenter)+(1|Set)+(1|Order), family = "binomial", data = lucid1)) # AIC BIC logLik deviance df.resid # 274.0 310.3 -126.0 252.0 189 # # Scaled residuals: # Min 1Q Median 3Q Max # -2.2021 -0.8246 -0.5063 0.8283 2.0272 # # Random effects: # Groups Name Variance Std.Dev. # ID (Intercept) 1.069e-06 0.0010339 # Item (Intercept) 1.866e-01 0.4319268 # Experimenter (Intercept) 8.217e-07 0.0009065 # Set (Intercept) 2.608e-06 0.0016150 # Order (Intercept) 3.788e-07 0.0006154 # Number of obs: 200, groups: ID, 44; Item, 5; Experimenter, 2; Set, 2; Order, 2 # # Fixed effects: # Estimate Std. Error z value Pr(>|z|) # (Intercept) -0.77818 0.40973 -1.899 0.0575 . # EFB1 0.72580 0.32393 2.241 0.0251 * # Complements 0.05134 0.02412 2.128 0.0333 * # MV3 0.22562 0.27756 0.813 0.4163 # Thinkyes -0.57835 0.53352 -1.084 0.2784 # MV3:Thinkyes 0.58057 0.45517 1.275 0.2021