--- title: "Study 5 - original analysis" author: "Matthew Leitao" date: "2/15/2022" output: html_document: #code_folding: hide df_print: paged toc: yes toc_float: true toc_depth: 5 toc_collapse: False editor_options: chunk_output_type: inline --- ```{r setup, include=FALSE} knitr::opts_chunk$set(echo = TRUE) ``` ```{r loading packages} if(!require(haven)) {install.packages("haven")} if(!require(tidyverse)) {install.packages("tidyverse")} if(!require(psych)) {install.packages("psych")} if(!require(ggplot2)) {install.packages("ggplot2")} if(!require(ggdist)) {install.packages("ggdist")} if(!require(gghalves)) {install.packages("gghalves")} if(!require(PupillometryR)) {install.packages("PupillometryR")} if(!require(jmv)) {install.packages("jmv")} if(!require(plyr)) {install.packages("plyr")} if(!require(lme4)) {install.packages("lme4")} if(!require(lmerTest)) {install.packages("lmerTest")} if(!require(modelsummary)) {install.packages("modelsummary")} if(!require(MuMIn)) {install.packages("MuMIn")} if(!require(papaja)) {install.packages("papaja")} if(!require(reghelper)) {install.packages("reghelper")} if(!require(emmeans)) {install.packages("emmeans")} if(!require(lavaan)) {install.packages("lavaan")} if(!require(semPlot)) {install.packages("semPlot")} if(!require(bmlm)) {install.packages("bmlm")} if(!require(sjPlot)) {install.packages("sjPlot")} if(!require(sjmisc)) {install.packages("sjmisc")} if(!require(EMAtools)) {install.packages("EMAtools")} if(!require(report)) {install.packages("report")} if(!require(MplusAutomation)) {install.packages("MplusAutomation")} if(!require(rstatix)) {install.packages("rstatix")} ``` #Health Algorithm - Belief 1 # Health Algorithm - Belief 3 ## Hypotheses Hypothesis 1: People in the voiced conditions will on average rate that they trust health recommendations more than those people in the unvoiced conditions Hypothesis 1a: People will report that they trust the doctor more when the doctor is accompanied by a voice than when he is not accompanied by a voice. Hypothesis 2: We will find a significant within-subject effect of condition on people’s ratings of trust. Hypothesis 2a: Participants on average will rate that they trust the health recommendation of a human expert more than the recommendation given by a non-human-like algorithm. Hypothesis 2b Participants on average will rate that they trust the health recommendation of a human-like algorithm more than the recommendation given by a non-human-like algorithm. ## Planned Done - We will use a Mixed Effects ANCOVA with planned comparisons. Done - We initially will determine whether the between-subject Voiced/Unvoiced conditions are significantly different in general. Done - We will then do a planned comparison between people's ratings of trust in the voiced doctor condition and the unvoiced doctor's condition Done - We will also determine whether the within-subject conditions of the recommenders is significant in predicting trust We will then do two planned comparisons. Done - First comparing ratings of trust in the human expert condition to the non-humanized algorithm condition Done - Second comparing ratings of trust in the humanized algorithm condition to the non-humanized algorithm condition ## Exploritory Analysis Done - Evaluate the within-subject difference in trust between a humanized algorithm and a human expert. Not Done - We will also evaluate the difference in trust between the voiced humanized algorithm and the unvoiced human expert. Done - Evaluate the interaction effects between the between-subjects conditions of Voiced/Unvoiced and the within-subjects conditions of recommendation giver. Not Done - We intend to evaluate how ascribing human-like traits to the recommender impacts ratings of trust as well as whether these attributions of humanness mediate the effect of condition in predicting trust. Not Done - We intend to look into the relationship between a few our conditions and some additional variables in predicting trust Done - If participants report that they prefer the advice from an algorithm or a doctor Not Done - How gender identity, age, socio-economic status, and ethnicity may interact with the conditions Not Done - How human-like they see the algorithm (through a direct question) Not Done - How human-like they see the algorithm through Mind Perception Indices Not Done - How appropriate do participants believe the doctor/algorithm for making these types of health decisions? Not Done - Examine the differences in how much participants blame/praise the recommender for successes and failures of the recommendation based on the recommender ```{r} describe(Digital_Beliefs_3_Tight$Age) prop.table(table(Digital_Beliefs_3$Gender)) Digital_Beliefs_3_Tight %>% filter(!is.na(Age)) %>% summarise(Mean = mean(IsMale)) ``` ```{r} ggplot(Digital_Beliefs_3_Longer, aes(x = as.factor(Cond), y = TruBel_WS, fill = Cond, col = Cond, na.rm = TRUE)) + geom_boxplot() + facet_wrap(~as.factor(IsVoiced)) ``` #Planned Comparision ## Plots Digital_Belief_Voiced_Longer %>% filter(!is.na(Cond)) %>% ggplot(., aes(x = Cond, y = Trust_WS, fill = Cond, col = Cond, na.rm = TRUE)) + geom_flat_violin(position = position_nudge(x = .2, y = 0), alpha = .25, adjust = 4)+ geom_point(position=position_jitter(width = .15),size = 1, alpha = 1/3)+ scale_fill_manual(values = c("#AF1B3F", "#B0A61A", "#144252"))+ scale_colour_manual(values = c("#5B2C38", "#333221", "#1B3842"))+ guides(fill = FALSE, col = FALSE, scale = "none") + geom_boxplot(aes(x = as.numeric(Cond)+ 0.25, y = Trust_WS),outlier.shape = NA, alpha = 0.4, width = .1, colour = "BLACK") + stat_summary(fun.data = "mean_cl_normal", geom = "errorbar", width = 1/4, color = "black") + stat_summary(fun = "mean", geom = "point",shape = 23, stroke = 1, size = 1.5, color = "black", fill = "#FBC760") + ggtitle("Trust In Recommendation by Recommender - Within Subjects") + ylab('Trust In Recommendation')+ xlab('Recommender') + theme_apa() + scale_y_continuous(breaks=seq(-3,3,.5)) guides(fill = FALSE, col = FALSE, scale = "none") + ```{r} Digital_Beliefs_3_Longer$Cond <- factor( Digital_Beliefs_3_Longer$Cond, levels = c(1,2,3), labels =c("Humanlike AI", "Standard AI", "Doctor") ) Digital_Beliefs_3_Longer$Cond ``` ```{r} as.numeric(Digital_Beliefs_3_Longer$IsVoiced) Digital_Beliefs_3_Longer <- Digital_Beliefs_3_Longer %>% mutate( IsVoiced_Fix = case_when( IsVoiced == 1 ~ 0, IsVoiced == 2 ~ 1, TRUE ~ NA_real_ ) ) Digital_Beliefs_3_Longer$IsVoiced <- Digital_Beliefs_3_Longer$IsVoiced_Fix Digital_Beliefs_3_Longer$IsVoiced <- factor( Digital_Beliefs_3_Longer$IsVoiced, levels = c(0,1), labels =c("Written Only", "Written and Voiced") ) Digital_Beliefs_3_Longer$IsVoiced ``` ```{r} Digital_Beliefs_3_Longer %>% filter(!is.na(Cond)) %>% mutate(Cond = fct_reorder(Cond, TruBel_WS, .fun = "mean")) %>% ggplot(., aes(x = Cond, y = TruBel_WS, fill = Cond, col = Cond, na.rm = TRUE)) + geom_flat_violin(position = position_nudge(x = .2, y = 0), alpha = .75, adjust = 4)+ geom_point(position=position_jitter(width = .15),size = 2, alpha = 1/2, color = "black")+ geom_boxplot(position = position_nudge(x = .25, y = 0),outlier.shape = NA, alpha = 0.4, width = .1, colour = "BLACK") + scale_fill_manual(values = c("#F8737F", "#FCBB6D", "#475C7A"))+ scale_colour_manual(values = c("#F8737F", "#FCBB6D", "#475C7A"))+ guides(fill = FALSE, col = FALSE, scale = "none") + stat_summary(fun.data = "mean_cl_normal", geom = "errorbar", width = 1/4, size = 1.25, color = "white") + stat_summary(fun = "mean", geom = "point",shape = 23, stroke = 1.25, size = 4, color = "white") + ggtitle("Trust/Belief In Recommendation by Recommender") + ylab('Trust/Belif In Recommendation') + xlab('Recommender') + theme_apa() + theme(plot.title = element_text(face = "bold", size = 16), axis.title = element_text(face = "bold", size = 14), panel.background = element_rect("transparent"), plot.background = element_rect("transparent", color = NA), legend.background = element_blank(), legend.box.background = element_rect("transparent", colour = 'white'), axis.text = element_text(colour = "white", size = 14, face = "bold"), axis.ticks = element_line(colour = "white"), axis.line = element_line(colour = "white")) + scale_y_continuous(breaks=seq(-3,3,.5)) ggsave("Study 5 - Trust and Belief Means - Within Subjects.png", width = 12, height = 8) ``` ```{r} Digital_Beliefs_3_Longer %>% filter(!is.na(Cond)) %>% mutate(Cond = fct_reorder(Cond, TruBel_WS, .fun = "mean")) %>% ggplot(., aes(x = as.factor(IsVoiced), y = TruBel_WS, fill = IsVoiced, col = IsVoiced, na.rm = TRUE)) + geom_flat_violin(position = position_nudge(x = .2, y = 0), alpha = .75, adjust = 4)+ geom_point(position=position_jitter(width = .15),size = 2, alpha = 1/2, color = "black")+ facet_wrap(~ Cond) + geom_boxplot(position = position_nudge(x = .25, y = 0),outlier.shape = NA, alpha = 0.4, width = .1, colour = "BLACK") + scale_fill_manual(values = c("#F2CC8C", "#325D79"))+ scale_colour_manual(values = c("#F2CC8C", "#325D79"))+ guides(fill = FALSE, col = FALSE, scale = "none") + stat_summary(fun.data = "mean_cl_normal", geom = "errorbar", width = 1/4, size = 1.5, color = "white") + stat_summary(fun = "mean", geom = "point",shape = 23, stroke = 1.5, size = 4, color = "white") + ggtitle("Trust/Belief In Recommendation by Recommender") + ylab('Trust/Belief In Recommendation') + xlab('Condition') + theme_apa() + theme(plot.title = element_text(face = "bold", size = 16), axis.title = element_text(face = "bold", size = 14), panel.background = element_rect("transparent"), plot.background = element_rect("transparent", color = NA), legend.background = element_blank(), legend.box.background = element_rect("transparent", colour = 'white'), axis.text = element_text(colour = "white", size = 14, face = "bold"), axis.ticks = element_line(colour = "white"), axis.line = element_line(colour = "white"), strip.text = element_text(colour = "white", size = 14, face = "bold")) + scale_y_continuous(breaks=seq(-3,3,.5)) ggsave("Study 5 - Trust and Belief Means - Within Subjects - Voice - Written.png", width = 12, height = 8) ``` ## Voiced vs UnVoiced - Doctor ```{r} ttestIS( formula = Cond_3.Trust_WS + Cond_3.Believe_WS + Cond_3.TruBel_WS ~ IsVoiced, data = Digital_Beliefs_3_Tight, vars = vars(Cond_3.Trust_WS, Cond_3.Believe_WS, Cond_3.TruBel_WS), meanDiff = TRUE, effectSize = TRUE, desc = TRUE, plots = TRUE) ``` ```{r} tidy(t.test(Cond_3.TruBel_WS ~ IsVoiced, data = Digital_Beliefs_3_Tight)) ``` ```{r} report(t.test(Cond_3.TruBel_WS ~ IsVoiced, data = Digital_Beliefs_3_Tight)) ``` ## Voiced vs UnVoiced - All ```{r} ttestIS( formula = Trust_WS + Believe_WS + TruBel_WS ~ IsVoiced, data = Digital_Beliefs_3_Longer, vars = vars(Trust_WS, Believe_WS, TruBel_WS), meanDiff = TRUE, effectSize = TRUE, desc = TRUE, plots = TRUE) ``` ## Voiced vs UnVoiced - Human-Like ```{r} ttestIS( formula = Cond_1.Trust_WS + Cond_1.Believe_WS + Cond_1.TruBel_WS ~ IsVoiced, data = Digital_Beliefs_3_Tight, vars = vars(Cond_1.Trust_WS, Cond_1.Believe_WS, Cond_1.TruBel_WS), meanDiff = TRUE, effectSize = TRUE, desc = TRUE, plots = TRUE) ``` ```{r} tidy(t.test(Cond_1.TruBel_WS ~ IsVoiced, data = Digital_Beliefs_3_Tight)) ``` ```{r} report(t.test(Cond_1.TruBel_WS ~ IsVoiced, data = Digital_Beliefs_3_Tight)) ``` ## Voiced vs UnVoiced - Non-Human-Like ```{r} ttestIS( formula = Cond_2.Trust_WS + Cond_2.Believe_WS + Cond_2.TruBel_WS ~ IsVoiced, data = Digital_Beliefs_3_Tight, vars = vars(Cond_2.Trust_WS, Cond_2.Believe_WS, Cond_2.TruBel_WS), meanDiff = TRUE, effectSize = TRUE, desc = TRUE, plots = TRUE) ``` ```{r} tidy(t.test(Cond_2.TruBel_WS ~ IsVoiced, data = Digital_Beliefs_3_Tight)) ``` ```{r} report(t.test(Cond_2.TruBel_WS ~ IsVoiced, data = Digital_Beliefs_3_Tight)) ``` ```{r} HealBelief.3 <- aov(TruBel_WS + Believe_WS + TruBel_WS ~ IsVoiced + Cond, data = Digital_Beliefs_3_Longer) summary(HealBelief.3) ``` ```{r} HealBelief.3 <- anova_test(data = Digital_Beliefs_3_Longer, dv = TruBel_WS, wid = ResponseId, between = IsVoiced, within= Cond, detailed = TRUE, effect.size = "pes") get_anova_table(HealBelief.3) ``` ```{r} anova_summary(Mixed_Effects_Anova_3.1, detailed = TRUE) ``` ### Non-Human-Like vs Doctor ## Fix to Paired Samples T-Test ```{r} Digital_Beliefs_3_Longer %>% filter(Cond != 1) %>% ttestIS( formula = Trust_WS + Believe_WS + TruBel_WS ~ Cond, vars = vars(Trust_WS, Believe_WS, TruBel_WS), meanDiff = TRUE, effectSize = TRUE, desc = TRUE, plots = TRUE) ``` Need to try and run this with voiced and unvoiced condition only for a robustness check ```{r} Digital_Beliefs_3_Tight %>% dplyr::filter(IsVoiced == 0) %>% jmv::ttestPS( data = ., pairs = list( list( i1="Cond_1.TruBel_WS", i2="Cond_3.TruBel_WS"), list( i1="Cond_1.TruBel_WS", i2="Cond_2.TruBel_WS"), list( i1="Cond_3.TruBel_WS", i2="Cond_2.TruBel_WS")), effectSize = TRUE, plots = TRUE) ``` ```{r} Digital_Beliefs_3 %>% dplyr::select(Cond_1.TruBel_WS, Cond_2.TruBel_WS, Cond_3.TruBel_WS) %>% describe() ``` ```{r} Digital_Beliefs_3_Tight %>% dplyr::filter(IsVoiced == 1) %>% jmv::ttestPS( data = ., pairs = list( list( i1="Cond_1.TruBel_WS", i2="Cond_3.TruBel_WS"), list( i1="Cond_1.TruBel_WS", i2="Cond_2.TruBel_WS"), list( i1="Cond_3.TruBel_WS", i2="Cond_2.TruBel_WS")), effectSize = TRUE, plots = TRUE) ``` ```{r} jmv::ttestPS( data = Digital_Beliefs_3_Tight, pairs = list( list( i1="Cond_1.TruBel_WS", i2="Cond_3.TruBel_WS"), list( i1="Cond_1.TruBel_WS", i2="Cond_2.TruBel_WS"), list( i1="Cond_3.TruBel_WS", i2="Cond_2.TruBel_WS")), effectSize = TRUE, plots = TRUE) ``` ### Human-Like vs Non-Human-Like ```{r} Digital_Beliefs_3_Longer %>% filter(Cond != 3) %>% ttestIS( formula = Trust_WS + Believe_WS + TruBel_WS ~ Cond, vars = vars(Trust_WS, Believe_WS, TruBel_WS), meanDiff = TRUE, effectSize = TRUE, desc = TRUE, plots = TRUE) ``` ### Human-Like vs Doctor ```{r} Digital_Beliefs_3_Longer %>% filter(Cond != 2) %>% ttestIS( formula = Trust_WS + Believe_WS + TruBel_WS ~ Cond, vars = vars(Trust_WS, Believe_WS, TruBel_WS), meanDiff = TRUE, effectSize = TRUE, desc = TRUE, plots = TRUE) ``` ```{r} Digital_Beliefs_3_Tight %>% t.test(Cond_1.TruBel_WS ~ IsVoiced) ``` # Mixed Effects ANOVA ## Pre-Registered ```{r} Mixed_Effects_Anova_3.1 <- aov(TruBel_WS ~ as.factor(Cond) + as.factor(IsVoiced) + Error(ResponseId/IsVoiced), data = Digital_Beliefs_3_Longer) summary(Mixed_Effects_Anova_3.1) anova_summary(Mixed_Effects_Anova_3.1, effect.size = "pes")3 ``` ```{r} report(Mixed_Effects_Anova_3.1) ``` ```{r} Mixed_Effects_Anova_3.1 <- aov(TruBel_WS ~ as.factor(Cond) + as.factor(IsVoiced) + Error(ResponseId/IsVoiced), data = Digital_Beliefs_3_Longer) anova_summary(Mixed_Effects_Anova_3.1) ``` ```{r} Mixed_Effects_Anova_3.1 <- aov(TruBel_WS ~ as.factor(Cond) * as.factor(IsVoiced) + Error(ResponseId/IsVoiced), data = Digital_Beliefs_3_Longer) summary(Mixed_Effects_Anova_3.1) anova_summary(Mixed_Effects_Anova_3.1, detailed = TRUE, effect.size = "pes") ``` ```{r} report(Mixed_Effects_Anova_3.1) ``` ```{r} model.tables(Mixed_Effects_Anova_3.1, "means") ``` ```{r} Mixed_Effects_Anova_3 <- lmer(TruBel_WS ~ as.factor(Cond) + as.factor(IsVoiced) + (1|ResponseId), data = Digital_Beliefs_3_Longer) summary(Mixed_Effects_Anova_3) ``` ```{r} tab_model(Mixed_Effects_Anova_3, show.stat = TRUE, show.std = TRUE, show.df = TRUE) ``` ```{r} lme.dscore(Mixed_Effects_Anova_3, data = Digital_Beliefs_3_Longer, type = "lme4") ``` ```{r} r.squaredGLMM(Mixed_Effects_Anova_3) ``` ```{r} report(Mixed_Effects_Anova_3) ``` ## Exploritory Interaction ```{r} Mixed_Effects_Anova_3_Exp <- lmer(TruBel_WS ~ as.factor(Cond) * as.factor(IsVoiced) + (1|ResponseId), data = Digital_Beliefs_3_Longer) summary(Mixed_Effects_Anova_3_Exp) ``` ```{r} r.squaredGLMM(Mixed_Effects_Anova_3_Exp) ``` ```{r} lme.dscore(Mixed_Effects_Anova_3_Exp, data = Digital_Beliefs_3_Longer, type = "lme4") ``` ```{r} anova(Mixed_Effects_Anova_3, Mixed_Effects_Anova_3_Exp) ``` ### Full Robust Model + Praise_WS + Blame_WS + Mind_Perception + Person Mean Centering ```{r} Digital_Beliefs_3_Longer$Mind_Perception_PM <- pmean(Digital_Beliefs_3_Longer$ResponseId, Digital_Beliefs_3_Longer$Mind_Perception) Digital_Beliefs_3_Longer$Mind_Perception_PMC <- pcenter(Digital_Beliefs_3_Longer$ResponseId, Digital_Beliefs_3_Longer$Mind_Perception) Digital_Beliefs_3_Longer$Warm_WS_PM <- pmean(Digital_Beliefs_3_Longer$ResponseId, Digital_Beliefs_3_Longer$Warm_WS) Digital_Beliefs_3_Longer$Warm_WS_PMC <- pcenter(Digital_Beliefs_3_Longer$ResponseId, Digital_Beliefs_3_Longer$Warm_WS) Digital_Beliefs_3_Longer$Competent_WS_PM <- pmean(Digital_Beliefs_3_Longer$ResponseId, Digital_Beliefs_3_Longer$Competent_WS) Digital_Beliefs_3_Longer$Competent_WS_PMC <- pcenter(Digital_Beliefs_3_Longer$ResponseId, Digital_Beliefs_3_Longer$Competent_WS) ``` ### Mind Perception Model ```{r} Mixed_Effects_Anova_3_MP <- lmer(TruBel_WS ~ as.factor(Cond) + as.factor(IsVoiced) + Mind_Perception + (1|ResponseId), data = Digital_Beliefs_3_Longer) summary(Mixed_Effects_Anova_3_MP) ``` ```{r} r.squaredGLMM(Mixed_Effects_Anova_3_MP) ``` ```{r} tab_model(Mixed_Effects_Anova_3_MP, show.std = TRUE) ``` ```{r} lme.dscore(Mixed_Effects_Anova_3_MP, data = Digital_Beliefs_3_Longer, type = "lme4") ``` ```{r} lm.beta.lmer(Mixed_Effects_Anova_3_MP) ``` ```{r} report(Mixed_Effects_Anova_3_MP) ``` ```{r} ```{r} Mixed_Effects_Anova_3_MP <- lmer(TruBel_WS ~ as.factor(Cond) + as.factor(IsVoiced) + Mind_Perception_Alt + (1|ResponseId), data = Digital_Beliefs_3_Longer) summary(Mixed_Effects_Anova_3_MP) ``` ```{r} tab_model(Mixed_Effects_Anova_3_MP, show.std = TRUE) ``` #### Warmth ```{r} Mixed_Effects_Anova_3_Cov <- lmer(TruBel_WS ~ as.factor(Cond) * as.factor(IsVoiced) + Mind_Perception + Warm_WS + Competent_WS + Age + IsMale + (1|ResponseId), data = Digital_Beliefs_3_Longer) summary(Mixed_Effects_Anova_3_Cov) ``` ```{r} r.squaredGLMM(Mixed_Effects_Anova_3_Cov) ``` ```{r} tab_model(Mixed_Effects_Anova_3_Cov, show.std = TRUE) ``` ```{r} lme.dscore(Mixed_Effects_Anova_3_Cov, data = Digital_Beliefs_3_Longer, type = "lme4") ``` ```{r} lm.beta.lmer(Mixed_Effects_Anova_3_Cov) ``` ```{r} report(Mixed_Effects_Anova_3_Cov) ``` ```{r} anova(Mixed_Effects_Anova_3_MP, Mixed_Effects_Anova_3_Cov) ``` ## Post Hoc ```{r} emmeans(Mixed_Effects_Anova_3_Exp, list(pairwise ~ Cond + IsVoiced), adjust = "tukey") ``` ```{r} lm.beta(Repeated_Measures) ``` ### Relgiousity ```{r} ReligiousMind <- lmer(Mind_Perception ~ Religious + (1|ResponseId), data = Digital_Beliefs_3_Longer) summary(ReligiousMind) ``` ```{r} tab_model(ReligiousMind) ``` ```{r} lme.dscore(ReligiousMind, data = Digital_Beliefs_3_Longer, type = "lme4") ``` ```{r} ReligiousMind <- lmer(Humanlike_WS ~ Religious + (1|ResponseId), data = Digital_Beliefs_3_Longer) summary(ReligiousMind) ``` ```{r} tab_model(ReligiousMind) ``` ```{r} lme.dscore(ReligiousMind, data = Digital_Beliefs_3_Longer, type = "lme4") ``` ### Mediation using Process for R #### Creating Variables ```{r} Digital_Beliefs_3_Tight <- Digital_Beliefs_3_Tight %>% mutate( Trubel_DocHum = Cond_3.TruBel_WS - Cond_1.TruBel_WS, MindPercept_DocHum = Cond_3.Mind_Perception - Cond_1.Mind_Perception, avgMindPercept_DocHum = (Cond_3.Mind_Perception + Cond_1.Mind_Perception)/2, gmavgMindPercept_DocHum = mean(avgMindPercept_DocHum, na.rm = TRUE), gmcavgMindPercept_DocHum = avgMindPercept_DocHum - gmavgMindPercept_DocHum, Const = 1 ) ``` ### Attempt to use Process The paths should look like path c = y ~ x path a = M ~ x path b = y ~ M path c' = y ~ x (with path b) path d = y ~ M+ ```{r} process(data = Digital_Beliefs_3_Tight, y = "Trubel_DocHum", x = "Const", m = "MindPercept_DocHum", model = 4, boot = 10000) ``` There doesn't work because x is a constant which makes me wonder how they ran it using their data. ```{r} DocHumPath <- 'Trubel_DocHum ~ Const Trubel_DocHum ~ MindPercept_DocHum Trubel_DocHum ~ gmavgMindPercept_DocHum MindPercept_DocHum ~ Const' DocHumPathAnalysis <- sem(model = DocHumPath, data = Digital_Beliefs_3_Tight, estimator = "ML") summary(PAModel, standardized = TRUE, fit.measures = TRUE, rsq = TRUE) ``` This also doesn't seem to work because there is no variation to the constant that Moytoya and Hayes mentioned using, the same error I receive when I try to use the processes. There something happening in their process code that I am unsure about. ### MLM based Mediation This is based on Vuorre & Bolger 2017. This will theoretically allow me to determine the mediation for multiple groups #### Base/Path C- Trubel ~ Condition ```{r} Mixed_Effects_Anova_3_Exp <- lmer(TruBel_WS ~ as.factor(Cond) + as.factor(IsVoiced) + (1|ResponseId), data = Digital_Beliefs_3_Longer) summary(Mixed_Effects_Anova_3_Exp) ``` #### A Path - Mind_Perception ~ Condition ```{r} Mixed_Effects_Anova_3_A.Path <- lmer(Mind_Perception ~ as.factor(Cond) + (1|ResponseId), data = Digital_Beliefs_3_Longer) summary(Mixed_Effects_Anova_3_A.Path) ``` #### Path B&C' - Trubel ~ Condition + Mind_Peception ```{r} Mixed_Effects_Anova_3_Path.BC <- lmer(TruBel_WS ~ as.factor(Cond) + as.factor(IsVoiced) + Mind_Perception + (1|ResponseId), data = Digital_Beliefs_3_Longer) summary(Mixed_Effects_Anova_3_Path.BC) ``` #### Indirect Effect ```{r} # Condition 1 - 2 -0.86898*0.40143 # Condition 1 - 3 0.08506*0.40143 ``` ### Bayes Model This is also from Vuorre & Bolger (2017) ```{r} bmlm.mindpercept <- Digital_Beliefs_3_Longer %>% drop_na () %>% mlm(d = ., id = "ResponseId", x = "Cond", m = "Mind_Perception", y = "TruBel_WS") ``` ```{r} mlm_summary(bmlm.mindpercept) ``` ```{r} mlm_path_plot(bmlm.mindpercept) ``` ```{r} -.83*.38 ``` ## Mplus Export File ```{r} Digital_Beliefs_3_Longer <- Digital_Beliefs_3_Longer %>% mutate( IsDoctor = case_when( Cond == 3 ~ 1, Cond == 1 ~ 0, Cond == 2 ~ 0, TRUE ~ NA_real_ ), IsNon = case_when( Cond == 2 ~ 1, Cond == 1 ~ 0, Cond == 3 ~ 0, TRUE ~ NA_real_ ) ) ``` ### Only Humanized/NonHumanized Algorithm ```{r} Digital_Beliefs_3_Longer %>% dplyr::select(ID, TruBel_WS, Mind_Perception, Cond, Warm_WS, Competent_WS, IsNon, IsDoctor) %>% drop_na() %>% write_csv(., "MPLus_Multilevel_Mediation_Study3_AlgOnly.csv") ``` ## Mplus Automation ### Mind Perception #### Humanlike AI vs Standardized AI ```{r} MPlus_MindPercept_Study3 <- Digital_Beliefs_3_Longer %>% dplyr::select(ID, TruBel_WS, Mind_Perception, Cond, Warm_WS, Competent_WS, IsDoctor, IsNon) %>% drop_na() #%>% # write_csv(., "MPLus_Multilevel_Mediation_Study3.csv") describe(MPlus_MindPercept_Study3) ``` ```{r} Mediation_Mind <- mplusObject( TITLE = "MIND PERCEPTION MEDIATION", VARIABLE = " USEVARIABLES ARE ID TruBel_WS IsNon IsDoctor; CLUSTER IS ID;", ANALYSIS = "TYPE IS TWOLEVEL RANDOM;", MODEL = "%WITHIN% TruBel_WS IsNon IsDoctor; TruBel_WS ON IsNon; TruBel_WS ON IsDoctor; %BETWEEN% TruBel_WS IsNon IsDoctor; TruBel_WS ON IsNon; TruBel_WS ON IsDoctor;", OUTPUT = "SAMPSTAT STANDARDIZED(CLUSTER) STDYX TECH1 TECH8 CINTERVAL;", usevariables = colnames(MPlus_MindPercept_Study3), rdata = MPlus_MindPercept_Study3) Mediation_Mind_Output <- mplusModeler(Mediation_Mind, dataout = "Mind_R.dat", modelout = "Mind_R.inp", check = TRUE, run = TRUE, hashfilename = FALSE) Test_Output_Mplus <- readModels("Mind_R.out") Test_Output_Mplus$parameters ``` ```{r} Mediation_Mind <- mplusObject( TITLE = "MIND PERCEPTION MEDIATION", VARIABLE = " USEVARIABLES ARE ID TruBel_WS Mind_Perception IsDoctor IsNon; CLUSTER IS ID;", ANALYSIS = "TYPE IS TWOLEVEL RANDOM;", MODEL = "%WITHIN% TruBel_WS Mind_Perception IsDoctor IsNon; TruBel_WS ON IsDoctor; TruBel_WS ON IsNon; Mind_Perception ON IsDoctor(a); Mind_Perception ON IsNon(x); TruBel_WS ON Mind_Perception(b); %BETWEEN% TruBel_WS Mind_Perception IsDoctor IsNon; TruBel_WS ON Mind_Perception; TruBel_WS ON IsDoctor; TruBel_WS ON IsNon; MODEL CONSTRAINT: NEW(indb); indb=a*b; NEW(inxy); inxy=x*b;", OUTPUT = "SAMPSTAT STANDARDIZED(CLUSTER) STDYX TECH1 TECH8 CINTERVAL;", usevariables = colnames(MPlus_MindPercept_Study3), rdata = MPlus_MindPercept_Study3) Mediation_Mind_Output <- mplusModeler(Mediation_Mind, dataout = "Mind_R.dat", modelout = "Mind_R.inp", check = TRUE, run = TRUE, hashfilename = FALSE) Test_Output_Mplus <- readModels("Mind_R.out") Test_Output_Mplus$parameters Test_Output_Mplus$data_summary ``` #### Humanlike AI vs Doctor ```{r} MPlus_MindPercept_Study3 <- Digital_Beliefs_3_Longer %>% filter(Cond != 2) %>% dplyr::select(ID, TruBel_WS, Mind_Perception, Cond, Warm_WS, Competent_WS, IsDoctor, IsNon) %>% drop_na() #%>% # write_csv(., "MPLus_Multilevel_Mediation_Study3.csv") describe(MPlus_MindPercept_Study3) ``` ```{r} Mediation_Mind <- mplusObject( TITLE = "MIND PERCEPTION MEDIATION", VARIABLE = " USEVARIABLES ARE ID TruBel_WS IsDoctor; CLUSTER IS ID;", ANALYSIS = "TYPE IS TWOLEVEL RANDOM;", MODEL = "%WITHIN% TruBel_WS IsDoctor; TruBel_WS ON IsDoctor; %BETWEEN% TruBel_WS IsDoctor; TruBel_WS ON IsDoctor;", OUTPUT = "SAMPSTAT STANDARDIZED(CLUSTER) STDYX TECH1 TECH8 CINTERVAL;", usevariables = colnames(MPlus_MindPercept_Study3), rdata = MPlus_MindPercept_Study3) Mediation_Mind_Output <- mplusModeler(Mediation_Mind, dataout = "Mind_R.dat", modelout = "Mind_R.inp", check = TRUE, run = TRUE, hashfilename = FALSE) Test_Output_Mplus <- readModels("Mind_R.out") Test_Output_Mplus$parameters ``` ```{r} Mediation_Mind <- mplusObject( TITLE = "MIND PERCEPTION MEDIATION", VARIABLE = " USEVARIABLES ARE ID TruBel_WS Mind_Perception IsDoctor; CLUSTER IS ID;", ANALYSIS = "TYPE IS TWOLEVEL RANDOM;", MODEL = "%WITHIN% TruBel_WS IsDoctor; TruBel_WS ON IsDoctor; Mind_Perception ON IsDoctor(a); TruBel_WS ON Mind_Perception(b); %BETWEEN% TruBel_WS Mind_Perception IsDoctor; Mind_Perception ON IsDoctor; TruBel_WS ON Mind_Perception; TruBel_WS ON IsDoctor; MODEL CONSTRAINT: NEW(indb); indb=a*b;", OUTPUT = "SAMPSTAT STANDARDIZED(CLUSTER) STDYX TECH1 TECH8 CINTERVAL;", usevariables = colnames(MPlus_MindPercept_Study3), rdata = MPlus_MindPercept_Study3) Mediation_Mind_Output <- mplusModeler(Mediation_Mind, dataout = "Mind_R.dat", modelout = "Mind_R.inp", check = TRUE, run = TRUE, hashfilename = FALSE) Test_Output_Mplus <- readModels("Mind_R.out") Test_Output_Mplus$parameters Test_Output_Mplus$data_summary ``` #### Standardized AI vs Doctor ```{r} MPlus_MindPercept_Study3 <- Digital_Beliefs_3_Longer %>% filter(Cond != 1) %>% dplyr::select(ID, TruBel_WS, Mind_Perception, Cond, Warm_WS, Competent_WS, IsDoctor, IsNon) %>% drop_na() #%>% # write_csv(., "MPLus_Multilevel_Mediation_Study3.csv") describe(MPlus_MindPercept_Study3) ``` ```{r} Mediation_Mind <- mplusObject( TITLE = "MIND PERCEPTION MEDIATION", VARIABLE = " USEVARIABLES ARE ID TruBel_WS IsDoctor; CLUSTER IS ID;", ANALYSIS = "TYPE IS TWOLEVEL RANDOM;", MODEL = "%WITHIN% TruBel_WS IsDoctor; TruBel_WS ON IsDoctor; %BETWEEN% TruBel_WS IsDoctor; TruBel_WS ON IsDoctor;", OUTPUT = "SAMPSTAT STANDARDIZED(CLUSTER) STDYX TECH1 TECH8 CINTERVAL;", usevariables = colnames(MPlus_MindPercept_Study3), rdata = MPlus_MindPercept_Study3) Mediation_Mind_Output <- mplusModeler(Mediation_Mind, dataout = "Mind_R.dat", modelout = "Mind_R.inp", check = TRUE, run = TRUE, hashfilename = FALSE) Test_Output_Mplus <- readModels("Mind_R.out") Test_Output_Mplus$parameters ``` ```{r} Mediation_Mind <- mplusObject( TITLE = "MIND PERCEPTION MEDIATION", VARIABLE = " USEVARIABLES ARE ID TruBel_WS Mind_Perception IsDoctor; CLUSTER IS ID;", ANALYSIS = "TYPE IS TWOLEVEL RANDOM;", MODEL = "%WITHIN% TruBel_WS IsDoctor; TruBel_WS ON IsDoctor; Mind_Perception ON IsDoctor(a); TruBel_WS ON Mind_Perception(b); %BETWEEN% TruBel_WS Mind_Perception IsDoctor; Mind_Perception ON IsDoctor; TruBel_WS ON Mind_Perception; TruBel_WS ON IsDoctor; MODEL CONSTRAINT: NEW(indb); indb=a*b;", OUTPUT = "SAMPSTAT STANDARDIZED(CLUSTER) STDYX TECH1 TECH8 CINTERVAL;", usevariables = colnames(MPlus_MindPercept_Study3), rdata = MPlus_MindPercept_Study3) Mediation_Mind_Output <- mplusModeler(Mediation_Mind, dataout = "Mind_R.dat", modelout = "Mind_R.inp", check = TRUE, run = TRUE, hashfilename = FALSE) Test_Output_Mplus <- readModels("Mind_R.out") Test_Output_Mplus$parameters Test_Output_Mplus$data_summary ``` #### Stop ```{r} Mediation_Mind <- mplusObject( TITLE = "MIND PERCEPTION MEDIATION", VARIABLE = " USEVARIABLES ARE ID TruBel_WS Mind_Perception Cond; CLUSTER IS ID;", ANALYSIS = "TYPE IS TWOLEVEL RANDOM;", MODEL = "%WITHIN% TruBel_WS Cond; TruBel_WS ON Cond; Mind_Perception ON Cond(a); TruBel_WS ON Mind_Perception(b); %BETWEEN% Cond Mind_Perception Cond; Mind_Perception ON Cond; TruBel_WS ON Mind_Perception; TruBel_WS ON Cond; MODEL CONSTRAINT: NEW(indb); indb=a*b;", OUTPUT = "SAMPSTAT STANDARDIZED(CLUSTER) STDYX TECH1 TECH8 CINTERVAL;", usevariables = colnames(MPlus_MindPercept_Study3), rdata = MPlus_MindPercept_Study3) Mediation_Mind_Output <- mplusModeler(Mediation_Mind, dataout = "Mind_R.dat", modelout = "Mind_R.inp", check = TRUE, run = TRUE, hashfilename = FALSE) Test_Output_Mplus <- readModels("Mind_R.out") Test_Output_Mplus$parameters ``` ## Clustering with Lavaan ```{r} Mediation_Path <- ' #Direct Effect TruBel_WS ~ c*Cond #Mediator Mind_Perception ~ a*Cond TruBel_WS ~ b*Mind_Perception #Indirect Effect (a*b) ab :=a*b #total effect total := c + (a*b) ' PathModel <- sem(Mediation_Path, data = Digital_Beliefs_3_Longer, group = "ResponseId") summary(PathModel, standardized = TRUE, fit.measures = TRUE, rsq = TRUE) ``` ```{r} fitmeasures(PathModel, c("chisq", "cfi", "rmsea", "df")) ``` ```{r} semPaths(PathModel, what = 'std', whatLabel ='std', residuals = TRUE, rotation = 2,sizeMan=6, edge.label.cex=.8) ``` ## Doc First #```{r} pairwise.t.test(Digital_Belief_Combined$WS_Non_DocFirst_Trust, Digital_Belief_Combined$Study, p.adjust.method = "none") pairwise.t.test(Digital_Belief_Combined$WS_Non_DocFirst_Believe, Digital_Belief_Combined$Study, p.adjust.method = "none") #``` #```{r} ttestIS( formula = WS_Non_DocFirst_Trust + WS_Non_DocFirst_Believe ~ Study, data = Digital_Belief_Combined, vars = vars(WS_Non_DocFirst_Trust, WS_Non_DocFirst_Believe), meanDiff = TRUE, effectSize = TRUE, desc = TRUE, plots = TRUE) #```