Section 1
Accuracy improves
mdlA <- lmer(unsigned_pe_1 ~ exam + (1 | cohort_class / pt_id),
data = df.retest)
summary(mdlA)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: unsigned_pe_1 ~ exam + (1 | cohort_class/pt_id)
## Data: df.retest
##
## REML criterion at convergence: 17843.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.8676 -0.7264 -0.1637 0.5384 4.6664
##
## Random effects:
## Groups Name Variance Std.Dev.
## pt_id:cohort_class (Intercept) 6.6924 2.5870
## cohort_class (Intercept) 0.5313 0.7289
## Residual 55.4286 7.4450
## Number of obs: 2565, groups: pt_id:cohort_class, 873; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 12.8841 0.5543 3.9620 23.245 2.2e-05 ***
## exam -1.0641 0.1256 1962.7886 -8.475 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## exam -0.561
PE -> Updating (no cov)
mdlC <- lmer(next_update_1 ~ pe_1 + (1 + pe_1 | cohort_class / pt_id),
data = df.retest)
## boundary (singular) fit: see ?isSingular
# AIC(mdlC)
summary(mdlC)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: next_update_1 ~ pe_1 + (1 + pe_1 | cohort_class/pt_id)
## Data: df.retest
##
## REML criterion at convergence: 12680.8
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.9073 -0.4996 0.0076 0.5284 3.6268
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## pt_id:cohort_class (Intercept) 5.742e-02 0.23963
## pe_1 7.644e-02 0.27647 1.00
## cohort_class (Intercept) 3.566e+00 1.88836
## pe_1 6.968e-03 0.08347 -1.00
## Residual 2.250e+02 15.00129
## Number of obs: 1526, groups: pt_id:cohort_class, 696; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 0.6302 1.1635 2.0272 0.542 0.642
## pe_1 0.3096 0.0586 2.2242 5.283 0.027 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## pe_1 -0.769
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see ?isSingular
PE -> Updating (cov)
mdlA <- lmer(next_update_1 ~ pe_1 + delta_grade + (1 + pe_1 | cohort_class / pt_id),
data = df.retest)
## boundary (singular) fit: see ?isSingular
## Warning: Model failed to converge with 1 negative eigenvalue: -8.1e+02
summary(mdlA)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: next_update_1 ~ pe_1 + delta_grade + (1 + pe_1 | cohort_class/pt_id)
## Data: df.retest
##
## REML criterion at convergence: 11568.5
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.4035 -0.5586 -0.0129 0.5784 3.7440
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## pt_id:cohort_class (Intercept) 0.000e+00 0.00000
## pe_1 6.404e-02 0.25307 NaN
## cohort_class (Intercept) 1.928e-01 0.43907
## pe_1 2.531e-03 0.05031 -1.00
## Residual 1.047e+02 10.23092
## Number of obs: 1526, groups: pt_id:cohort_class, 696; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) -0.42898 0.37476 2.25619 -1.145 0.35919
## pe_1 0.68031 0.03924 2.10171 17.335 0.00265 **
## delta_grade 0.72445 0.01774 1368.95375 40.835 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) pe_1
## pe_1 -0.517
## delta_grade -0.044 0.239
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see ?isSingular
PE Valence Interaction
md <- lmer(next_update_1 ~ unsigned_pe_1_flip*pe_1_sign_f_NAzero + delta_grade + (1 | cohort_class / pt_id), data = df.retest)
summary(md)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula:
## next_update_1 ~ unsigned_pe_1_flip * pe_1_sign_f_NAzero + delta_grade +
## (1 | cohort_class/pt_id)
## Data: df.retest
##
## REML criterion at convergence: 11416.5
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.0166 -0.5395 0.0024 0.5785 3.5152
##
## Random effects:
## Groups Name Variance Std.Dev.
## pt_id:cohort_class (Intercept) 9.7804 3.1274
## cohort_class (Intercept) 0.4964 0.7046
## Residual 106.2767 10.3091
## Number of obs: 1503, groups: pt_id:cohort_class, 696; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value
## (Intercept) -1.83842 0.62991 3.74586 -2.919
## unsigned_pe_1_flip 0.74214 0.03679 1497.72828 20.174
## pe_1_sign_f_NAzero -0.21158 0.46120 1488.62315 -0.459
## delta_grade 0.72607 0.01782 920.04470 40.748
## unsigned_pe_1_flip:pe_1_sign_f_NAzero 0.14668 0.03596 1489.93035 4.079
## Pr(>|t|)
## (Intercept) 0.0469 *
## unsigned_pe_1_flip < 2e-16 ***
## pe_1_sign_f_NAzero 0.6465
## delta_grade < 2e-16 ***
## unsigned_pe_1_flip:pe_1_sign_f_NAzero 4.77e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) un__1_ p_1___ dlt_gr
## unsgnd_p_1_ 0.005
## p_1_sgn__NA -0.013 -0.771
## delta_grade 0.040 0.260 -0.023
## u__1_:_1___ -0.585 0.034 -0.022 -0.137
# updating rates still larger for positive PEs
plot_model(md, type = "pred", terms = c("unsigned_pe_1_flip","pe_1_sign_f_NAzero"))
pos rate
B = 0.91
md_pos <- lmer(next_update_1 ~ unsigned_pe_1 + delta_grade + (1 | cohort_class / pt_id), data = df.retest[which(df.retest$pe_1_sign_f_NAzero == "1"),])
## boundary (singular) fit: see ?isSingular
summary(md_pos)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: next_update_1 ~ unsigned_pe_1 + delta_grade + (1 | cohort_class/pt_id)
## Data: df.retest[which(df.retest$pe_1_sign_f_NAzero == "1"), ]
##
## REML criterion at convergence: 5837.1
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.7830 -0.5294 -0.0095 0.5291 3.4035
##
## Random effects:
## Groups Name Variance Std.Dev.
## pt_id:cohort_class (Intercept) 17.97 4.239
## cohort_class (Intercept) 0.00 0.000
## Residual 95.06 9.750
## Number of obs: 772, groups: pt_id:cohort_class, 469; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) -1.99860 0.65579 562.69097 -3.048 0.00241 **
## unsigned_pe_1 0.91140 0.05164 768.99853 17.651 < 2e-16 ***
## delta_grade 0.82571 0.02523 692.54879 32.731 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) uns__1
## unsignd_p_1 -0.763
## delta_grade 0.058 0.126
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see ?isSingular
neg rate
B = -0.52
md_neg <- lmer(next_update_1 ~ unsigned_pe_1 + delta_grade + (1 | cohort_class / pt_id), data = df.retest[which(df.retest$pe_1_sign_f_NAzero == "-1"),])
summary(md_neg)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: next_update_1 ~ unsigned_pe_1 + delta_grade + (1 | cohort_class/pt_id)
## Data: df.retest[which(df.retest$pe_1_sign_f_NAzero == "-1"), ]
##
## REML criterion at convergence: 5544.2
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.7767 -0.5233 0.0144 0.5959 2.7901
##
## Random effects:
## Groups Name Variance Std.Dev.
## pt_id:cohort_class (Intercept) 18.175 4.263
## cohort_class (Intercept) 1.171 1.082
## Residual 96.765 9.837
## Number of obs: 731, groups: pt_id:cohort_class, 490; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) -1.75137 0.92174 3.91616 -1.90 0.132
## unsigned_pe_1 -0.52733 0.05216 726.40500 -10.11 <2e-16 ***
## delta_grade 0.63619 0.02478 601.68075 25.68 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) uns__1
## unsignd_p_1 -0.544
## delta_grade 0.049 -0.385
Section 2
NE & Grades
summary(lmer(grade ~ neurot + (1 | cohort_class / pt_id),
data = df.retest))
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: grade ~ neurot + (1 | cohort_class/pt_id)
## Data: df.retest
##
## REML criterion at convergence: 21026.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.4604 -0.5707 0.1165 0.6120 2.3976
##
## Random effects:
## Groups Name Variance Std.Dev.
## pt_id:cohort_class (Intercept) 154.774 12.441
## cohort_class (Intercept) 2.675 1.635
## Residual 142.705 11.946
## Number of obs: 2543, groups: pt_id:cohort_class, 865; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 73.74553 1.74524 15.07561 42.255 <2e-16 ***
## neurot -0.10023 0.08884 812.90837 -1.128 0.26
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## neurot -0.787
NE & Initial Expectations
summary(lmer(pred_1 ~ neurot + (1 | cohort_class / pt_id),
data = df.retest))
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: pred_1 ~ neurot + (1 | cohort_class/pt_id)
## Data: df.retest
##
## REML criterion at convergence: 20553.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.8675 -0.3970 0.1398 0.5476 3.2732
##
## Random effects:
## Groups Name Variance Std.Dev.
## pt_id:cohort_class (Intercept) 116.153 10.777
## cohort_class (Intercept) 9.511 3.084
## Residual 122.224 11.055
## Number of obs: 2543, groups: pt_id:cohort_class, 865; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 77.0892 2.2013 4.1778 35.020 2.56e-06 ***
## neurot -0.2371 0.0784 779.9783 -3.024 0.00258 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## neurot -0.549
NE & Initial PEs
summary(lmer(pe_1 ~ neurot + exam + pe_1_lag1 + grade_lag1 + (1 | cohort_class / pt_id),
data = df.retest))
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula:
## pe_1 ~ neurot + exam + pe_1_lag1 + grade_lag1 + (1 | cohort_class/pt_id)
## Data: df.retest
##
## REML criterion at convergence: 11724.5
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.9504 -0.5355 0.0248 0.5614 3.6325
##
## Random effects:
## Groups Name Variance Std.Dev.
## pt_id:cohort_class (Intercept) 36.7125 6.0591
## cohort_class (Intercept) 0.5005 0.7075
## Residual 103.6687 10.1818
## Number of obs: 1515, groups: pt_id:cohort_class, 690; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) -10.14825 2.24355 362.35135 -4.523 8.26e-06 ***
## neurot 0.20398 0.06519 314.19837 3.129 0.00192 **
## exam 0.04706 0.29691 649.95383 0.159 0.87411
## pe_1_lag1 -0.04551 0.02840 1429.91167 -1.602 0.10928
## grade_lag1 0.10039 0.02301 1282.20515 4.362 1.39e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) neurot exam p_1_l1
## neurot -0.521
## exam -0.308 -0.013
## pe_1_lag1 0.490 -0.108 0.004
## grade_lag1 -0.761 0.098 -0.115 -0.571
NE & Accuracy
summary(lmer(unsigned_pe_1 ~ neurot + exam + unsigned_lag_pe_1 +(1 | cohort_class / pt_id),
data = df.retest))
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula:
## unsigned_pe_1 ~ neurot + exam + unsigned_lag_pe_1 + (1 | cohort_class/pt_id)
## Data: df.retest
##
## REML criterion at convergence: 10377
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.9819 -0.7156 -0.1982 0.5155 4.8612
##
## Random effects:
## Groups Name Variance Std.Dev.
## pt_id:cohort_class (Intercept) 2.0110 1.418
## cohort_class (Intercept) 0.8949 0.946
## Residual 52.8024 7.267
## Number of obs: 1515, groups: pt_id:cohort_class, 690; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 8.73248 1.05036 19.27931 8.314 8.44e-08 ***
## neurot 0.10460 0.03536 286.16455 2.959 0.003349 **
## exam -0.68699 0.20530 940.65575 -3.346 0.000851 ***
## unsigned_lag_pe_1 0.08553 0.02439 1503.29177 3.506 0.000468 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) neurot exam
## neurot -0.497
## exam -0.619 -0.010
## unsgnd_l__1 -0.272 -0.076 0.125
NE -> Updating
md <- lmer(next_update_1 ~ pe_1*neurot + delta_grade + (1 + neurot | cohort_class / pt_id), data = df.retest)
## boundary (singular) fit: see ?isSingular
summary(md)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula:
## next_update_1 ~ pe_1 * neurot + delta_grade + (1 + neurot | cohort_class/pt_id)
## Data: df.retest
##
## REML criterion at convergence: 11516.8
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.9571 -0.5488 -0.0221 0.5797 3.6035
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## pt_id:cohort_class (Intercept) 7.190e+00 2.68148
## neurot 1.122e-01 0.33491 -1.00
## cohort_class (Intercept) 1.905e+00 1.38014
## neurot 4.874e-03 0.06982 -1.00
## Residual 1.070e+02 10.34575
## Number of obs: 1515, groups: pt_id:cohort_class, 690; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 1.590e+00 1.185e+00 2.139e+00 1.342 0.30442
## pe_1 5.724e-01 6.485e-02 1.259e+03 8.827 < 2e-16 ***
## neurot -1.280e-01 6.880e-02 2.390e+00 -1.861 0.18264
## delta_grade 7.352e-01 1.752e-02 1.157e+03 41.968 < 2e-16 ***
## pe_1:neurot 9.832e-03 3.794e-03 1.143e+03 2.592 0.00967 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) pe_1 neurot dlt_gr
## pe_1 0.085
## neurot -0.959 -0.073
## delta_grade -0.002 0.169 -0.014
## pe_1:neurot -0.065 -0.932 0.050 -0.026
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see ?isSingular
Overlearning from small PEs
md <- lmer(next_uPE_1 ~ PE1_size*neurot*PE1_sign + exam + (1 + neurot | cohort_class / pt_id), df.retest,
control = lmerControl(check.nobs.vs.nRE="ignore"))
## boundary (singular) fit: see ?isSingular
summary(md)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: next_uPE_1 ~ PE1_size * neurot * PE1_sign + exam + (1 + neurot |
## cohort_class/pt_id)
## Data: df.retest
## Control: lmerControl(check.nobs.vs.nRE = "ignore")
##
## REML criterion at convergence: 10213.1
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.5898 -0.7218 -0.1829 0.5052 4.5241
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## pt_id:cohort_class (Intercept) 4.210e-01 0.64881
## neurot 2.357e-02 0.15351 -1.00
## cohort_class (Intercept) 2.202e+00 1.48402
## neurot 8.958e-04 0.02993 -1.00
## Residual 5.120e+01 7.15539
## Number of obs: 1492, groups: pt_id:cohort_class, 690; cohort_class, 3
##
## Fixed effects:
## Estimate Std. Error df t value
## (Intercept) 11.14075 1.45442 7.66903 7.660
## PE1_sizeSmall -4.38434 1.57250 1400.65577 -2.788
## neurot 0.01997 0.07286 46.84745 0.274
## PE1_sign+ -1.18000 1.70062 1423.13873 -0.694
## exam -0.66164 0.20545 1133.69850 -3.220
## PE1_sizeSmall:neurot 0.24492 0.09890 1309.81474 2.476
## PE1_sizeSmall:PE1_sign+ 1.55371 2.29932 1408.57987 0.676
## neurot:PE1_sign+ 0.02015 0.10171 1302.83185 0.198
## PE1_sizeSmall:neurot:PE1_sign+ -0.13546 0.14004 1328.75453 -0.967
## Pr(>|t|)
## (Intercept) 7.47e-05 ***
## PE1_sizeSmall 0.00537 **
## neurot 0.78524
## PE1_sign+ 0.48788
## exam 0.00132 **
## PE1_sizeSmall:neurot 0.01339 *
## PE1_sizeSmall:PE1_sign+ 0.49933
## neurot:PE1_sign+ 0.84301
## PE1_sizeSmall:neurot:PE1_sign+ 0.33356
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) PE1_sS neurot PE1_s+ exam PE1_sS: PE1_S:P n:PE1_
## PE1_sizSmll -0.526
## neurot -0.832 0.625
## PE1_sign+ -0.489 0.461 0.582
## exam -0.262 -0.016 0.001 -0.012
## PE1_szSmll: 0.498 -0.937 -0.673 -0.432 -0.003
## PE1_S:PE1_+ 0.368 -0.686 -0.427 -0.737 -0.021 0.644
## nrt:PE1_sg+ 0.486 -0.454 -0.658 -0.939 -0.003 0.490 0.693
## PE1_S::PE1_ -0.352 0.664 0.474 0.680 0.009 -0.709 -0.941 -0.724
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see ?isSingular
plot_model(md, type = "pred", terms = c("PE1_size", "neurot")) +
ylim(c(6,13)) +
ggtitle(NULL)
## Scale for 'y' is already present. Adding another scale for 'y', which will
## replace the existing scale.
plot_model(md, type = "pred", terms = c("PE1_sign", "neurot")) +
ylim(c(7,13)) +
ggtitle(NULL)
## Scale for 'y' is already present. Adding another scale for 'y', which will
## replace the existing scale.
