setwd("~/Documents/Dropbox/Research/Bernd")
p<-read.csv ("manipulating_TIF.csv", header=T, sep=",")
#Including only people who reported their attitudes twice (pre- and post-facts)
p<-subset(p, Rcars1_1 != "NA" | terror_concern_1_1 != "NA" | meat1_1 != "NA")
#Cars 1 pre-fact attitudes; cars 2 is post-fact attitudes. Both are the average of willingness to ride in and willingness to buy driverless cars.
p$cars1<-(p$Rcars1_1+p$Rcars1_2)/2
p$cars2<-(p$Rcars2_1+p$Rcars2_2)/2
#Attitudes are more positive post- vs. pre-fact.
t.test(p$cars1, p$cars2)##
## Welch Two Sample t-test
##
## data: p$cars1 and p$cars2
## t = -3.0076, df = 901.74, p-value = 0.002707
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -10.974271 -2.307335
## sample estimates:
## mean of x mean of y
## 48.50110 55.14191#Creating the "attitude change" variable (carsdiff). If this is positive, it means attitudes towards the car improved after learning the facts. If negative, attitudes worsened.
p$carsdiff<-(p$cars2 - p$cars1)
summary(p$carsdiff)## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
## -49.000 0.000 2.000 6.606 10.500 99.000 230#Same process for terror and meat variables.
p$terror1<-(p$terror_concern_1_1+p$terror_die_1_1)/2
p$terror2<-(p$terror_concern_2_1+p$terror_die_2_1)/2
#People are less concerned about terrorism post-facts.
t.test(p$terror1, p$terror2)##
## Welch Two Sample t-test
##
## data: p$terror1 and p$terror2
## t = 8.5065, df = 906.52, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 9.042574 14.466479
## sample estimates:
## mean of x mean of y
## 35.80947 24.05495p$terrordiff<-(p$terror1 - p$terror2)
summary(p$terrordiff)## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
## -50.00 1.50 7.75 11.98 19.12 74.00 229#People are more positive about lab grown meat post-facts.
t.test(p$meat1_1, p$meat2_1)##
## Welch Two Sample t-test
##
## data: p$meat1_1 and p$meat2_1
## t = -3.5491, df = 896.22, p-value = 0.0004066
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -12.048514 -3.468046
## sample estimates:
## mean of x mean of y
## 51.73392 59.49220p$meatdiff<-(p$meat2_1 - p$meat1_1)
summary(p$meatdiff)## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
## -50.000 0.000 2.000 7.795 11.250 100.000 233#Creating subsets of the data based on manipulated condition (high or low trust in feelings), in order to do t-tests to check if attitude change varied by condition. Also excluding people who failed the basic attention checks for the manipulations.
hiT<-subset(p, hiTIF==1)
hiT<-subset(p, ch1==4 & ch2==4 & ch3==4)
loT<-subset(p, loTIF==1)
loT<-subset(p, ch4==1 & ch5==1&ch6==2)
t.test(hiT$carsdiff, loT$carsdiff)##
## Welch Two Sample t-test
##
## data: hiT$carsdiff and loT$carsdiff
## t = 1.2761, df = 385.14, p-value = 0.2027
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -0.8971882 4.2155152
## sample estimates:
## mean of x mean of y
## 7.745370 6.086207t.test(hiT$meatdiff, loT$meatdiff)##
## Welch Two Sample t-test
##
## data: hiT$meatdiff and loT$meatdiff
## t = 0.67237, df = 403.29, p-value = 0.5017
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -2.136438 4.357528
## sample estimates:
## mean of x mean of y
## 8.535545 7.425000t.test(hiT$terrordiff, loT$terrordiff)##
## Welch Two Sample t-test
##
## data: hiT$terrordiff and loT$terrordiff
## t = -1.0325, df = 414.57, p-value = 0.3024
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -4.319796 1.344564
## sample estimates:
## mean of x mean of y
## 11.24884 12.73645#The manipulation didn't affect attitude change, nor did it affect attitudes pre- or post-facts.
t.test(hiT$terror2, loT$terror2)##
## Welch Two Sample t-test
##
## data: hiT$terror2 and loT$terror2
## t = 1.3554, df = 417.99, p-value = 0.176
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -1.200360 6.532604
## sample estimates:
## mean of x mean of y
## 24.55093 21.88480t.test(hiT$cars2, loT$cars2)##
## Welch Two Sample t-test
##
## data: hiT$cars2 and loT$cars2
## t = 0.38344, df = 416.91, p-value = 0.7016
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -5.177131 7.686413
## sample estimates:
## mean of x mean of y
## 55.44676 54.19212t.test(hiT$meat2, loT$meat2)##
## Welch Two Sample t-test
##
## data: hiT$meat2 and loT$meat2
## t = -0.15469, df = 406.84, p-value = 0.8771
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -6.771223 5.783299
## sample estimates:
## mean of x mean of y
## 59.81604 60.31000t.test(hiT$terror1, loT$terror1)##
## Welch Two Sample t-test
##
## data: hiT$terror1 and loT$terror1
## t = 0.61454, df = 408.53, p-value = 0.5392
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -2.755917 5.262666
## sample estimates:
## mean of x mean of y
## 35.65581 34.40244t.test(hiT$cars1, loT$cars1)##
## Welch Two Sample t-test
##
## data: hiT$cars1 and loT$cars1
## t = -0.14135, df = 418.92, p-value = 0.8877
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -6.980528 6.043935
## sample estimates:
## mean of x mean of y
## 47.63761 48.10591t.test(hiT$meat1_1, loT$meat1_1)##
## Welch Two Sample t-test
##
## data: hiT$meat1_1 and loT$meat1_1
## t = -0.40678, df = 408.08, p-value = 0.6844
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -7.926594 5.208525
## sample estimates:
## mean of x mean of y
## 51.51659 52.87562#Reliance on feelings does correlate with attitude change for all 3 targets. More reliance on feelings, less attitude change. Even stronger correlations between initial attitudes and reliance on feelings.
p$carfeel<-(p$carsmod_1 + (102 - p$carsmod_2))/2
cor.test(p$carfeel, p$carsdiff)##
## Pearson's product-moment correlation
##
## data: p$carfeel and p$carsdiff
## t = -1.889, df = 446, p-value = 0.05954
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.180247739 0.003583656
## sample estimates:
## cor
## -0.08909069cor.test(p$carfeel, p$cars1)##
## Pearson's product-moment correlation
##
## data: p$carfeel and p$cars1
## t = -12.998, df = 448, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.5873172 -0.4527504
## sample estimates:
## cor
## -0.5232884p$meatfeel<-(p$meatmod_1+(102-p$meatmod_2))/2
cor.test(p$meatfeel, p$meatdiff)##
## Pearson's product-moment correlation
##
## data: p$meatfeel and p$meatdiff
## t = -1.7129, df = 443, p-value = 0.08743
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.17276768 0.01193329
## sample estimates:
## cor
## -0.08111352cor.test(p$meatfeel, p$meat1_1)##
## Pearson's product-moment correlation
##
## data: p$meatfeel and p$meat1_1
## t = -11.984, df = 446, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.5605501 -0.4200992
## sample estimates:
## cor
## -0.4935356p$terrorfeel<-(p$terrormod_1+(102-p$terrormod_2))/2
cor.test(p$terrorfeel, p$terrordiff)##
## Pearson's product-moment correlation
##
## data: p$terrorfeel and p$terrordiff
## t = -3.1369, df = 444, p-value = 0.001821
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.23686211 -0.05514967
## sample estimates:
## cor
## -0.1472481cor.test(p$terrorfeel, p$terror1)##
## Pearson's product-moment correlation
##
## data: p$terrorfeel and p$terror1
## t = 6.3338, df = 446, p-value = 5.848e-10
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.1999488 0.3700678
## sample estimates:
## cor
## 0.2872721p$manip[p$loTIF==1]<-0
p$manip[p$hiTIF==1]<-1
#Does the manipulation interact with people's "trait" tendency to rely on their feelings when making decisions (either in general or with regards to new technologies)?
p$techmode<-(p$techmode_1 + (102-p$techmode_2))/2
p$generalmode<-(p$generalmode_1 + (102-p$generalmode_2))/2
#Marginal interactions for reliance on feelings "in general"; significant interactions for reliance on feelings for tech.
summary(lm(terrordiff ~ manip * generalmode, data=p))##
## Call:
## lm(formula = terrordiff ~ manip * generalmode, data = p)
##
## Residuals:
## Min 1Q Median 3Q Max
## -59.883 -9.680 -3.690 7.356 60.363
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 9.80778 2.44108 4.018 6.89e-05 ***
## manip -4.98293 3.70604 -1.345 0.179
## generalmode 0.07158 0.05392 1.327 0.185
## manip:generalmode 0.06894 0.07918 0.871 0.384
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 14.61 on 448 degrees of freedom
## (229 observations deleted due to missingness)
## Multiple R-squared: 0.01978, Adjusted R-squared: 0.01322
## F-statistic: 3.014 on 3 and 448 DF, p-value: 0.0298summary(lm(carsdiff ~ manip * generalmode, data=p))##
## Call:
## lm(formula = carsdiff ~ manip * generalmode, data = p)
##
## Residuals:
## Min 1Q Median 3Q Max
## -54.294 -6.669 -4.227 3.301 90.523
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 5.91470 2.56019 2.310 0.0213 *
## manip -4.07110 3.73601 -1.090 0.2764
## generalmode -0.01218 0.05667 -0.215 0.8299
## manip:generalmode 0.14353 0.07978 1.799 0.0727 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 14.37 on 447 degrees of freedom
## (230 observations deleted due to missingness)
## Multiple R-squared: 0.01901, Adjusted R-squared: 0.01242
## F-statistic: 2.887 on 3 and 447 DF, p-value: 0.03532summary(lm(meatdiff ~ manip * generalmode, data=p))##
## Call:
## lm(formula = meatdiff ~ manip * generalmode, data = p)
##
## Residuals:
## Min 1Q Median 3Q Max
## -59.398 -7.894 -4.876 2.952 94.132
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 8.23812 3.36510 2.448 0.0147 *
## manip -7.38579 5.05636 -1.461 0.1448
## generalmode -0.02014 0.07350 -0.274 0.7842
## manip:generalmode 0.18193 0.10755 1.691 0.0914 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 18.13 on 444 degrees of freedom
## (233 observations deleted due to missingness)
## Multiple R-squared: 0.01017, Adjusted R-squared: 0.003477
## F-statistic: 1.52 on 3 and 444 DF, p-value: 0.2086summary(lm(terrordiff ~ manip * polimode_1, data=p))##
## Call:
## lm(formula = terrordiff ~ manip * polimode_1, data = p)
##
## Residuals:
## Min 1Q Median 3Q Max
## -60.961 -9.704 -4.082 7.959 58.971
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 9.69345 2.01047 4.821 1.96e-06 ***
## manip -0.98749 3.11021 -0.317 0.751
## polimode_1 0.06204 0.03536 1.754 0.080 .
## manip:polimode_1 -0.01693 0.05326 -0.318 0.751
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 14.67 on 448 degrees of freedom
## (229 observations deleted due to missingness)
## Multiple R-squared: 0.01269, Adjusted R-squared: 0.006079
## F-statistic: 1.92 on 3 and 448 DF, p-value: 0.1256summary(lm(carsdiff ~ manip * techmode, data=p))##
## Call:
## lm(formula = carsdiff ~ manip * techmode, data = p)
##
## Residuals:
## Min 1Q Median 3Q Max
## -53.370 -6.988 -4.142 3.574 90.422
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 7.99103 2.16541 3.690 0.000252 ***
## manip -4.32407 3.20706 -1.348 0.178246
## techmode -0.07100 0.05330 -1.332 0.183536
## manip:techmode 0.16730 0.07334 2.281 0.023016 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 14.37 on 447 degrees of freedom
## (230 observations deleted due to missingness)
## Multiple R-squared: 0.01881, Adjusted R-squared: 0.01222
## F-statistic: 2.856 on 3 and 447 DF, p-value: 0.03679summary(lm(meatdiff ~ manip * techmode, data=p))##
## Call:
## lm(formula = meatdiff ~ manip * techmode, data = p)
##
## Residuals:
## Min 1Q Median 3Q Max
## -55.537 -8.642 -4.162 3.895 90.851
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 11.99445 2.88482 4.158 3.86e-05 ***
## manip -10.68617 4.24326 -2.518 0.01214 *
## techmode -0.12661 0.07191 -1.761 0.07898 .
## manip:techmode 0.29343 0.09984 2.939 0.00346 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 18.04 on 444 degrees of freedom
## (233 observations deleted due to missingness)
## Multiple R-squared: 0.02018, Adjusted R-squared: 0.01356
## F-statistic: 3.048 on 3 and 444 DF, p-value: 0.02851#To break down the interactions, we can look at people who are above and below the mid point of the trait measure.
hiF<-subset(p, techmode>=50)
loF<-subset(p, techmode<50)
#People who already rely on their feelings a lot change their attitudes more in the "relying on feelings is good" condition than in the "relying on feelings is bad" condition. I guess because they aren't basing their attitudes on the facts we provide them but rather on their feelings?
summary(lm(carsdiff~manip, data=hiF))##
## Call:
## lm(formula = carsdiff ~ manip, data = hiF)
##
## Residuals:
## Min 1Q Median 3Q Max
## -51.359 -7.702 -2.359 3.383 91.298
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.359 2.163 1.091 0.2771
## manip 5.343 2.870 1.861 0.0647 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 17.3 on 146 degrees of freedom
## (64 observations deleted due to missingness)
## Multiple R-squared: 0.02318, Adjusted R-squared: 0.01649
## F-statistic: 3.465 on 1 and 146 DF, p-value: 0.0647summary(lm(carsdiff~manip, data=loF))##
## Call:
## lm(formula = carsdiff ~ manip, data = loF)
##
## Residuals:
## Min 1Q Median 3Q Max
## -27.363 -7.239 -4.615 3.385 84.637
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 6.615 1.005 6.583 2.05e-10 ***
## manip 1.248 1.468 0.850 0.396
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 12.75 on 301 degrees of freedom
## (166 observations deleted due to missingness)
## Multiple R-squared: 0.002395, Adjusted R-squared: -0.0009196
## F-statistic: 0.7225 on 1 and 301 DF, p-value: 0.396summary(lm(meatdiff~manip, data=hiF))##
## Call:
## lm(formula = meatdiff ~ manip, data = hiF)
##
## Residuals:
## Min 1Q Median 3Q Max
## -56.30 -10.95 -6.30 3.70 88.05
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 6.300 2.582 2.44 0.016 *
## manip 5.647 3.464 1.63 0.105
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 20 on 133 degrees of freedom
## (77 observations deleted due to missingness)
## Multiple R-squared: 0.01959, Adjusted R-squared: 0.01222
## F-statistic: 2.658 on 1 and 133 DF, p-value: 0.1054summary(lm(meatdiff~manip, data=loF))##
## Call:
## lm(formula = meatdiff ~ manip, data = loF)
##
## Residuals:
## Min 1Q Median 3Q Max
## -55.324 -7.770 -5.324 4.676 93.676
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 7.770 1.343 5.785 1.77e-08 ***
## manip -1.445 1.953 -0.740 0.46
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 17.25 on 311 degrees of freedom
## (156 observations deleted due to missingness)
## Multiple R-squared: 0.001758, Adjusted R-squared: -0.001452
## F-statistic: 0.5476 on 1 and 311 DF, p-value: 0.4599all <- reshape(p,
varying = c("carsdiff", "meatdiff", "terrordiff"),
v.names = "change",
timevar = "target",
times = c("car", "meat", "terror"),
direction = "long")
#Creating dummy variables.
all$meat[all$target=="meat"]<-1
all$meat[all$target!="meat"]<-0
all$car[all$target=="car"]<-1
all$car[all$target!="car"]<-0
all$terror[all$target=="terror"]<-1
all$terror[all$target!="terror"]<-0
#Using just manipulation to predict attitude change:
summary(lm(change ~ manip, data=all))##
## Call:
## lm(formula = change ~ manip, data = all)
##
## Residuals:
## Min 1Q Median 3Q Max
## -59.051 -8.547 -4.551 4.953 90.949
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 8.5471 0.6149 13.900 <2e-16 ***
## manip 0.5042 0.8719 0.578 0.563
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 16.02 on 1349 degrees of freedom
## (692 observations deleted due to missingness)
## Multiple R-squared: 0.0002479, Adjusted R-squared: -0.0004932
## F-statistic: 0.3345 on 1 and 1349 DF, p-value: 0.5631#Using manipulation and the dummy variables for targets.
summary(lm(change ~ manip + meat + car + terror, data=all))##
## Call:
## lm(formula = change ~ manip + meat + car + terror, data = all)
##
## Residuals:
## Min 1Q Median 3Q Max
## -62.249 -8.062 -4.839 4.938 92.128
##
## Coefficients: (1 not defined because of singularities)
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 11.7161 0.8593 13.634 < 2e-16 ***
## manip 0.5328 0.8634 0.617 0.537
## meat -4.1867 1.0578 -3.958 7.96e-05 ***
## car -5.3767 1.0561 -5.091 4.06e-07 ***
## terror NA NA NA NA
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 15.87 on 1347 degrees of freedom
## (692 observations deleted due to missingness)
## Multiple R-squared: 0.02103, Adjusted R-squared: 0.01885
## F-statistic: 9.648 on 3 and 1347 DF, p-value: 2.657e-06#Using manipulation, dummies, and their interactions.
summary(lm(change ~ manip + meat + car + terror + manip*meat + manip*car + manip*terror, data=all))##
## Call:
## lm(formula = change ~ manip + meat + car + terror + manip * meat +
## manip * car + manip * terror, data = all)
##
## Residuals:
## Min 1Q Median 3Q Max
## -61.152 -8.152 -4.904 4.785 91.785
##
## Coefficients: (2 not defined because of singularities)
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 12.784 1.048 12.200 < 2e-16 ***
## manip -1.631 1.492 -1.094 0.274351
## meat -5.406 1.488 -3.632 0.000292 ***
## car -7.379 1.488 -4.958 8.04e-07 ***
## terror NA NA NA NA
## manip:meat 2.469 2.114 1.168 0.243157
## manip:car 4.030 2.111 1.909 0.056448 .
## manip:terror NA NA NA NA
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 15.86 on 1345 degrees of freedom
## (692 observations deleted due to missingness)
## Multiple R-squared: 0.02373, Adjusted R-squared: 0.0201
## F-statistic: 6.537 on 5 and 1345 DF, p-value: 5.108e-06#Using manipulation and trait reliance on feelings and their interaction.
summary(lm(change ~ manip * generalmode, data=all))##
## Call:
## lm(formula = change ~ manip * generalmode, data = all)
##
## Residuals:
## Min 1Q Median 3Q Max
## -60.002 -8.560 -4.237 5.083 92.986
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 7.97062 1.62427 4.907 1.04e-06 ***
## manip -5.37118 2.42459 -2.215 0.0269 *
## generalmode 0.01371 0.03577 0.383 0.7017
## manip:generalmode 0.12871 0.05171 2.489 0.0129 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 15.95 on 1347 degrees of freedom
## (692 observations deleted due to missingness)
## Multiple R-squared: 0.01103, Adjusted R-squared: 0.008832
## F-statistic: 5.01 on 3 and 1347 DF, p-value: 0.001858#Breaking down that interaction, splitting into high and low reliance on feelings people.
summary(all$generalmode)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 1.00 32.50 46.00 43.75 55.00 86.50hiF<-subset(all, generalmode>=46)
loF<-subset(all, generalmode<46)
#Again, people who already rely on their feelings (hiF) a lot change their attitudes more in the "relying on feelings is good" condition than in the "relying on feelings is bad" condition.
summary(lm(change~manip, data=hiF))##
## Call:
## lm(formula = change ~ manip, data = hiF)
##
## Residuals:
## Min 1Q Median 3Q Max
## -60.224 -9.724 -4.815 5.685 88.776
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 8.3146 0.9647 8.619 <2e-16 ***
## manip 2.9097 1.3259 2.194 0.0285 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 17.28 on 680 degrees of freedom
## (353 observations deleted due to missingness)
## Multiple R-squared: 0.007032, Adjusted R-squared: 0.005572
## F-statistic: 4.816 on 1 and 680 DF, p-value: 0.02854hiFcond_hiFtrait<-subset(hiF, hiTIF==1)
loFcond_hiFtrait<-subset(hiF, loTIF==1)
t.test(hiFcond_hiFtrait$change, loFcond_hiFtrait$change)##
## Welch Two Sample t-test
##
## data: hiFcond_hiFtrait$change and loFcond_hiFtrait$change
## t = 2.2146, df = 678.96, p-value = 0.02712
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 0.330002 5.489468
## sample estimates:
## mean of x mean of y
## 11.224377 8.314642#The opposite is true for people who don't rely on their feelings: they change their attitudes more in the "relying on feelings is bad" condition.
summary(lm(change~manip, data=loF))##
## Call:
## lm(formula = change ~ manip, data = loF)
##
## Residuals:
## Min 1Q Median 3Q Max
## -56.529 -7.756 -4.256 4.244 93.471
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 8.7556 0.7643 11.456 <2e-16 ***
## manip -2.2266 1.1209 -1.986 0.0474 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 14.46 on 667 degrees of freedom
## (339 observations deleted due to missingness)
## Multiple R-squared: 0.005881, Adjusted R-squared: 0.004391
## F-statistic: 3.946 on 1 and 667 DF, p-value: 0.04739hiFcond_loFtrait<-subset(loF, hiTIF==1)
loFcond_loFtrait<-subset(loF, loTIF==1)
t.test(hiFcond_loFtrait$change, loFcond_loFtrait$change)##
## Welch Two Sample t-test
##
## data: hiFcond_loFtrait$change and loFcond_loFtrait$change
## t = -1.9802, df = 644.77, p-value = 0.04811
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -4.4346991 -0.0185963
## sample estimates:
## mean of x mean of y
## 6.528939 8.755587#Interaction between manipulation and political party.
summary(lm(change ~ manip * party, data=all))##
## Call:
## lm(formula = change ~ manip * party, data = all)
##
## Residuals:
## Min 1Q Median 3Q Max
## -60.176 -8.176 -4.676 5.073 91.573
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 10.314 2.099 4.913 1.01e-06 ***
## manip -5.636 2.962 -1.902 0.0573 .
## party -1.104 1.254 -0.880 0.3789
## manip:party 3.853 1.774 2.171 0.0301 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 16 on 1347 degrees of freedom
## (692 observations deleted due to missingness)
## Multiple R-squared: 0.004365, Adjusted R-squared: 0.002147
## F-statistic: 1.968 on 3 and 1347 DF, p-value: 0.1169#Democrats change their attitudes more in the "feelings are good" condition... I guess because their initial feelings are already positive? Republicans aren't affected by the manipulation at all.
rep<-subset(all, party==1)
dem<-subset(all, party==2)
summary(lm(change~manip, data=rep))##
## Call:
## lm(formula = change ~ manip, data = rep)
##
## Residuals:
## Min 1Q Median 3Q Max
## -59.210 -8.210 -4.427 5.573 91.573
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 9.2103 0.9867 9.335 <2e-16 ***
## manip -1.7831 1.3903 -1.282 0.2
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 16.24 on 544 degrees of freedom
## (279 observations deleted due to missingness)
## Multiple R-squared: 0.003014, Adjusted R-squared: 0.001182
## F-statistic: 1.645 on 1 and 544 DF, p-value: 0.2002summary(lm(change~manip, data=dem))##
## Call:
## lm(formula = change ~ manip, data = dem)
##
## Residuals:
## Min 1Q Median 3Q Max
## -60.176 -8.107 -5.176 4.824 89.824
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 8.107 0.784 10.340 <2e-16 ***
## manip 2.070 1.116 1.854 0.0641 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 15.84 on 803 degrees of freedom
## (413 observations deleted due to missingness)
## Multiple R-squared: 0.004262, Adjusted R-squared: 0.003022
## F-statistic: 3.437 on 1 and 803 DF, p-value: 0.06412#Democrats' initial attitudes are definitely more positive - we don't know if those attitudes are feelings-based, but they could be since there isn't a correlation between party and reliance on feelings.
t.test(rep$cars1, dem$cars1)##
## Welch Two Sample t-test
##
## data: rep$cars1 and dem$cars1
## t = -5.7503, df = 1101.8, p-value = 1.153e-08
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -14.251027 -6.999778
## sample estimates:
## mean of x mean of y
## 41.93353 52.55893t.test(rep$meat1, dem$meat1)##
## Welch Two Sample t-test
##
## data: rep$meat1 and dem$meat1
## t = -3.3011, df = 1172.6, p-value = 0.0009922
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -9.754334 -2.481780
## sample estimates:
## mean of x mean of y
## 48.09836 54.21642t.test(rep$terror1, dem$terror1)##
## Welch Two Sample t-test
##
## data: rep$terror1 and dem$terror1
## t = 10.55, df = 1242.2, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 9.522006 13.872595
## sample estimates:
## mean of x mean of y
## 42.5599 30.8626t.test(rep$generalmode, dem$generalmode)##
## Welch Two Sample t-test
##
## data: rep$generalmode and dem$generalmode
## t = 1.0634, df = 1791.5, p-value = 0.2878
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -0.6816367 2.2960478
## sample estimates:
## mean of x mean of y
## 44.23455 43.42734#However there is a correlation between social conservatism in particular and reliance on feelings...
cor.test(all$generalmode, all$poli2_2)##
## Pearson's product-moment correlation
##
## data: all$generalmode and all$poli2_2
## t = -3.662, df = 2041, p-value = 0.0002567
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.12372615 -0.03755702
## sample estimates:
## cor
## -0.08079254#In addition to the interaction between party and the reliance on feelings manipulatin, there is also an interaction between party and reliance on feelings (trait) - again, no effect of reliance on feelings for republicans, but for democrats more reliance on feelings leads to greater attitude change. This seems to suggest again that democrats just have more positive feelings towards these targets. So we want democrats to rely more on their feelings and republicans to rely less on their feelings... or just try to improve republican's feelings.
summary(lm(change ~ generalmode * party, data=all))##
## Call:
## lm(formula = change ~ generalmode * party, data = all)
##
## Residuals:
## Min 1Q Median 3Q Max
## -59.580 -8.394 -4.430 4.800 92.345
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 11.61380 4.15239 2.797 0.00523 **
## generalmode -0.09547 0.08808 -1.084 0.27864
## party -3.82975 2.46892 -1.551 0.12109
## generalmode:party 0.10742 0.05257 2.043 0.04122 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 15.95 on 1347 degrees of freedom
## (692 observations deleted due to missingness)
## Multiple R-squared: 0.01022, Adjusted R-squared: 0.008011
## F-statistic: 4.634 on 3 and 1347 DF, p-value: 0.003138summary(lm(change~generalmode, data=rep))##
## Call:
## lm(formula = change ~ generalmode, data = rep)
##
## Residuals:
## Min 1Q Median 3Q Max
## -58.513 -8.392 -4.481 5.057 90.612
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 7.78405 1.96526 3.961 8.46e-05 ***
## generalmode 0.01196 0.04160 0.287 0.774
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 16.27 on 544 degrees of freedom
## (279 observations deleted due to missingness)
## Multiple R-squared: 0.0001518, Adjusted R-squared: -0.001686
## F-statistic: 0.08261 on 1 and 544 DF, p-value: 0.7739summary(lm(change~generalmode, data=dem))##
## Call:
## lm(formula = change ~ generalmode, data = dem)
##
## Residuals:
## Min 1Q Median 3Q Max
## -59.580 -8.849 -4.163 4.622 92.345
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 3.95429 1.52195 2.598 0.009544 **
## generalmode 0.11938 0.03271 3.650 0.000279 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 15.74 on 803 degrees of freedom
## (413 observations deleted due to missingness)
## Multiple R-squared: 0.01632, Adjusted R-squared: 0.0151
## F-statistic: 13.32 on 1 and 803 DF, p-value: 0.00027912nd only