dat <- read.csv("~/Desktop/Aidan Results/ANALYZED JASP READY Aggregate Qualities of Memories of FIlm 06-02-26.csv")We expect that the Emotional Intensity of the memory will be positively correlated with Narrative Coherence, but Emotional Valence will not (Bohanek et al., 2005; Byrne et al., 2001).
mod.2b1=lm(Narrative.Coherence ~ Emotional.Intensity, data = dat)
summary(mod.2b1)##
## Call:
## lm(formula = Narrative.Coherence ~ Emotional.Intensity, data = dat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.9347 -0.8101 0.1153 0.8149 2.4396
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4.41057 0.22446 19.649 <2e-16 ***
## Emotional.Intensity 0.12484 0.05073 2.461 0.0146 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.163 on 235 degrees of freedom
## Multiple R-squared: 0.02513, Adjusted R-squared: 0.02098
## F-statistic: 6.057 on 1 and 235 DF, p-value: 0.01457Emotional intensity is positively correlated with narrative coherence.
mod.2b2=lm(Narrative.Coherence ~ Emotional.Valence,dat)
summary(mod.2b2)##
## Call:
## lm(formula = Narrative.Coherence ~ Emotional.Valence, data = dat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.77244 -0.93939 0.09366 0.82756 2.11654
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4.99024 0.16365 30.494 <2e-16 ***
## Emotional.Valence -0.01525 0.03713 -0.411 0.682
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.177 on 235 degrees of freedom
## Multiple R-squared: 0.0007176, Adjusted R-squared: -0.003535
## F-statistic: 0.1688 on 1 and 235 DF, p-value: 0.6816Emotional valence is not correlated with narrative coherence, but let’s not affirm the null. \(BF_{01}\) time!
library(BayesFactor)## Warning: package 'BayesFactor' was built under R version 4.5.2## Loading required package: coda## Loading required package: Matrix## ************
## Welcome to BayesFactor 0.9.12-4.8. If you have questions, please contact Richard Morey (richarddmorey@gmail.com).
##
## Type BFManual() to open the manual.
## ************# 1. Run the Bayesian linear regression
bf_model <- lmBF(Narrative.Coherence ~ Emotional.Valence, data = dat)
# 2. Extract the standard Bayes Factor (which is BF10)
bf10 <- extractBF(bf_model)$bf
# 3. Flip it to get BF01 (evidence FOR the null)
bf01 <- 1 / bf10
# 4. Print the result
print(paste("BF01 =", round(bf01, 3)))## [1] "BF01 = 6.499"The data are 6.5 times more likely under the null compared to the alternative. Thus, we can pretty confidently say that emotional valence does not correlate with narrative coherence.
If Emotional Intensity predicts Narrative Coherence and if Vividness predicts Narrative Coherence, we will then explore the extent to which Vividness mediates the relationship between Emotional Intensity and Narrative Coherence.
Emotional valence does not correlate with narrative coherence, so moving on.
Linear regressions were conducted to look at the relationship between Narrative Coherence and Vividness, Emotional Valence, and Emotional Intensity. These analyses were conducted independent of film grouping.
Since some of these were already done above, I’m assuming this is meant to be a multiple regression?
mod.2=lm(Narrative.Coherence ~ Vividness + Emotional.Valence + Emotional.Intensity, dat)
summary(mod.2)##
## Call:
## lm(formula = Narrative.Coherence ~ Vividness + Emotional.Valence +
## Emotional.Intensity, data = dat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.38401 -0.48695 0.04498 0.53169 2.62285
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.89552 0.25652 7.389 2.62e-12 ***
## Vividness 0.65625 0.04674 14.040 < 2e-16 ***
## Emotional.Valence -0.01359 0.02845 -0.478 0.633
## Emotional.Intensity -0.01476 0.04081 -0.362 0.718
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.8566 on 233 degrees of freedom
## Multiple R-squared: 0.4752, Adjusted R-squared: 0.4685
## F-statistic: 70.34 on 3 and 233 DF, p-value: < 2.2e-16Vividness is positively associated with narrative coherence, even after adjusting for emotional valence and intensity. Neither of the latter predict narrative coherence after adjusting for vividness. It’s noteworthy too that \(R^2\) was really low for the simple linear regressions but is now at 0.48, accounting for nearly half the variation in narrative coherence. That’s really high!
A simple linear regression was conducted with Vividness as the predictor and Narrative Coherence as the outcome. We looked for a significant model in which Vividness positively predicted Narrative Coherence.
mod.2a=lm(Narrative.Coherence~Vividness,dat)
summary(mod.2a)##
## Call:
## lm(formula = Narrative.Coherence ~ Vividness, data = dat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.33564 -0.52134 0.02722 0.50150 2.59008
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.80414 0.22187 8.132 2.44e-14 ***
## Vividness 0.65145 0.04476 14.555 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.8539 on 235 degrees of freedom
## Multiple R-squared: 0.4741, Adjusted R-squared: 0.4718
## F-statistic: 211.8 on 1 and 235 DF, p-value: < 2.2e-16Vivideness is indeed a significant predictor of narrative coherence.
A multiple linear regression was conducted with Emotional Valence and Emotional Intensity as predictors and Narrative Coherence as the outcome. We looked for a nonsignificant model for Emotional Valence and Narrative Coherence. Additionally, we looked for a significant model in which Emotional Intensity positively predicted Narrative Coherence. Finally, we looked for a nonsignificant interaction effect of both Emotional Valence and Emotional Intensity on Narrative Coherence.
summary(lm(Narrative.Coherence~Emotional.Valence+Emotional.Intensity,dat))##
## Call:
## lm(formula = Narrative.Coherence ~ Emotional.Valence + Emotional.Intensity,
## data = dat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3.00765 -0.83415 0.07349 0.83226 2.41667
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4.51143 0.23905 18.872 <2e-16 ***
## Emotional.Valence -0.04678 0.03843 -1.217 0.2247
## Emotional.Intensity 0.14438 0.05316 2.716 0.0071 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.161 on 234 degrees of freedom
## Multiple R-squared: 0.03126, Adjusted R-squared: 0.02298
## F-statistic: 3.776 on 2 and 234 DF, p-value: 0.02433Emotional valence does not predict narrative coherence once emotional intensity is accounted for, but emotional intensity does predict narrative coherence even while adjusting for emotional valence.
If significant correlations between Vividness and Narrative Coherence and Emotional Intensity and Narrative Coherence are observed, then a mediation analysis would be conducted with Vividness mediating the relationship between Emotional Intensity and Narrative Coherence. We would be looking for a significant indirect effect of Emotional Intensity on Narrative Coherence through Vividness as well as a significant total effect.
# Load needed package
library(lavaan)## Warning: package 'lavaan' was built under R version 4.5.2## This is lavaan 0.6-21
## lavaan is FREE software! Please report any bugs. # Define the mediation paths
med_model <- '
Vividness ~ a * Emotional.Intensity
Narrative.Coherence ~ b * Vividness + c * Emotional.Intensity
# Calculate indirect and total effects
indirect := a * b
total := c + (a * b)
'
# Fit the model with bootstrapping
fit_med <- sem(med_model, data = dat, se = "bootstrap", bootstrap = 1000)
# Print the summary results
summary(fit_med, standardized = TRUE, ci = TRUE)## lavaan 0.6-21 ended normally after 1 iteration
##
## Estimator ML
## Optimization method NLMINB
## Number of model parameters 5
##
## Number of observations 237
##
## Model Test User Model:
##
## Test statistic 0.000
## Degrees of freedom 0
##
## Parameter Estimates:
##
## Standard errors Bootstrap
## Number of requested bootstrap draws 1000
## Number of successful bootstrap draws 1000
##
## Regressions:
## Estimate Std.Err z-value P(>|z|) ci.lower ci.upper
## Vividness ~
## Emtnl.Intn (a) 0.221 0.053 4.170 0.000 0.124 0.329
## Narrative.Coherence ~
## Vividness (b) 0.658 0.045 14.519 0.000 0.565 0.746
## Emtnl.Intn (c) -0.021 0.041 -0.509 0.611 -0.101 0.059
## Std.lv Std.all
##
## 0.221 0.266
##
## 0.658 0.696
## -0.021 -0.026
##
## Variances:
## Estimate Std.Err z-value P(>|z|) ci.lower ci.upper
## .Vividness 1.427 0.117 12.200 0.000 1.188 1.638
## .Narrativ.Chrnc 0.722 0.070 10.254 0.000 0.579 0.849
## Std.lv Std.all
## 1.427 0.929
## 0.722 0.525
##
## Defined Parameters:
## Estimate Std.Err z-value P(>|z|) ci.lower ci.upper
## indirect 0.146 0.037 3.988 0.000 0.080 0.225
## total 0.125 0.054 2.299 0.021 0.020 0.236
## Std.lv Std.all
## 0.146 0.185
## 0.125 0.159The warning can be ignored. Both the indirect and total effects are significant. This could be read as, “Emotional intensity has an effect on narrative coherence and that effect is mediated by vividness.” There is no direct effect of emotional intensity on narrative coherence once vividness is accounted for. It can be called a “full mediation” because of that.