DecisionWeighting
LOADING DATA…
setwd("~/Dropbox/PROGRAMMING/R/WeightsExp/Pilot/data")
subjects = list.files(pattern="*.csv")
for (i in 1:length(subjects)){
assign(paste0("s", i), read.csv(subjects[i]))
}
setwd("~/Dropbox/PROGRAMMING/R/WeightsExp/Pilot")
#Initialize data frame to hold all subjects
total_NM <- NULL
total_M <- NULL
Subject <- NULL
###########################
# CLEAN DATA FRAMES
# (break into Multiplier and non-Multiplier)
###########################
#loop through and create data frames
for (i in 1:length(subjects)) {
temp <- eval(as.name(paste0("s", i)))
#remove extraneous rows
temp <- temp[-c(1:3), -c(1:21)]
#create non-multiplier data frame
temp_NM <- temp[c(1:100), c(1:10) ]
#create multiplier data frame
temp_M <- temp[c(102:401), c(1:15)]
temp_M <- temp_M[,-c(9:12)]
#renameSomeColumns
library(plyr)
temp_NM <- rename(temp_NM, c("numberEntry_3.keys"="acceptReject", "numberEntry_3.rt"="decisionTime"))
temp_M <- rename(temp_M, c("numberEntry_4.keys"="acceptReject", "numberEntry_4.rt"="decisionTime"))
#clean up acceptReject column
index <- c("['f']", "['j']")
values <- c(1, 0) #where 1 is accept and 0 is reject
temp_NM$acceptReject <- values[match(temp_NM$acceptReject, index)]
temp_M$acceptReject <- values[match(temp_M$acceptReject, index)]
#clean up Flip column
index <- c("True", 1, 2)
values <- c(1, 0, 0) #where 1 is accept and 0 is reject
temp_M$flip <- values[match(temp_M$flip, index)]
#make summed value a value to 2 decimal places
temp_NM$summedVal <- format(round(temp_NM$summedVal, 2), nsmall = 2)
temp_M$summedVal <- format(round(temp_M$summedVal, 2), nsmall = 2)
#make numeric
temp_NM$summedVal <- as.numeric(temp_NM$summedVal)
temp_M$summedVal <- as.numeric(temp_M$summedVal)
#remove brackets on RT column
temp_NM$decisionTime <- gsub("\\[|\\]", "", temp_NM$decisionTime)
temp_M$decisionTime <- gsub("\\[|\\]", "", temp_M$decisionTime)
#make Numeric
temp_NM$decisionTime <- as.numeric(temp_NM$decisionTime)
temp_M$decisionTime <- as.numeric(temp_M$decisionTime)
#########
# Save individual data frames as well (maybe a new loop after)
total_NM <- rbind(total_NM, temp_NM)
total_M <- rbind(total_M, temp_M)
}
#Make row numbering sequential
rownames(total_NM) <- 1:nrow(total_NM)
rownames(total_M) <- 1:nrow(total_M)
###CHECK THIS OUT###
#Replace NAs in flip column with 0s.
total_M[is.na(total_M)] <- 0
head(total_M)## Trial correct faceVal houseVal mult1House mult2Face summedVal earnings
## 1 1 1 0.44 -0.78 1 1 -0.34 32.88
## 2 2 0 -0.28 0.16 3 1 0.20 32.68
## 3 3 1 -0.10 -0.80 3 1 -2.50 32.68
## 4 4 1 0.84 -0.46 2 1 -0.08 32.68
## 5 5 1 0.82 0.40 1 1 1.22 33.90
## 6 6 1 0.38 0.20 1 1 0.58 34.48
## flip acceptReject decisionTime
## 1 0 0 8.916853
## 2 0 0 2.466852
## 3 0 0 4.816845
## 4 0 0 7.666896
## 5 0 1 2.633707
## 6 0 1 2.466849PRELIMINARY ANALYSIS
###########################
# ANALYSIS
###########################
#GROUP
#Percent Correct
mean(total_NM$correct)## [1] 0.8685714mean(total_M$correct)## [1] 0.8857143#Mean Response Time
mean(total_NM$decisionTime)## [1] 4.127875mean(total_M$decisionTime)## [1] 2.934391#clean outliers due to distraction
total_NM_clean<- total_NM[!(total_NM$decisionTime>10),]
total_M_clean<- total_M[!(total_M$decisionTime>10),]
#Cleaned Mean Response Time
mean(total_NM_clean$decisionTime)## [1] 3.073678mean(total_M_clean$decisionTime)## [1] 2.466951#SD Response Time
#Cleaned Mean Response Time
sd(total_NM_clean$decisionTime)## [1] 2.148276sd(total_M_clean$decisionTime)## [1] 1.700813#Mean response time with multipliers
mean(total_M_clean$decisionTime[total_M_clean$mult1House>1])## [1] 2.569192mean(total_M_clean$decisionTime[total_M_clean$mult2Face>1])## [1] 2.455839mean(total_M_clean$decisionTime[total_M_clean$mult1House>1])## [1] 2.569192mean(total_M_clean$decisionTime[total_M_clean$mult2Face>1])## [1] 2.455839#What about RT WITH TWO multipliers
mean(total_M_clean$decisionTime[(total_M_clean$mult2Face>1) & (total_M_clean$mult1House>1)])## [1] 2.809233#vs WITHOUT
mean(total_M_clean$decisionTime[(total_M_clean$mult2Face<2) & (total_M_clean$mult1House<2)])## [1] 2.462117#with only FACE multiplier
mean(total_M_clean$decisionTime[(total_M_clean$mult2Face>1) & (total_M_clean$mult1House==1)])## [1] 2.351807#with only HOUSE multiplier
mean(total_M_clean$decisionTime[(total_M_clean$mult2Face==1) & (total_M_clean$mult1House>1)])## [1] 2.494322PLOT STUFF
#RT
hist(total_M_clean$decisionTime, breaks = 50)
abline(v = mean(total_M_clean$decisionTime),
col = "royalblue",
lwd = 2)
#median line
abline(v = median(total_M_clean$decisionTime),
col = "red",
lwd = 2)
#LEGEND
legend(x = "topright",
c(as.expression(bquote(Mean == .(mean(total_M_clean$decisionTime)))), as.expression(bquote(Median == .(median(total_M_clean$decisionTime))))),
col = c("royalblue", "red"),
lwd = c(2, 2, 2))
# 3D Scatterplot with Coloring and Vertical Lines
# and Regression Plane
#RT vs. HVal, FVal
library(scatterplot3d)
s3d <-scatterplot3d(total_M_clean$decisionTime,total_M_clean$faceVal,total_M_clean$houseVal, pch=16, highlight.3d=TRUE,
type="h", main="3D Scatterplot")
fit <- lm(total_M_clean$decisionTime ~ total_M_clean$faceVal+total_M_clean$houseVal)
s3d$plane3d(fit)
#RT vs. Summed Value
library(ggplot2)
ggplot(total_M_clean, aes(x=summedVal, y=decisionTime)) +
geom_point(shape=1) + # Use hollow circles
geom_smooth() # Add a loess smoothed fit curve with confidence region## `geom_smooth()` using method = 'gam'
#Effect of Flips?
library(ggplot2)
p <- ggplot(total_M_clean, aes(factor(flip), decisionTime))
p + geom_boxplot()
Summed Value vs. Accept/Reject Decision
Logistic regression curve
plot(total_M_clean$summedVal, total_M_clean$acceptReject)
model <- glm(acceptReject ~ summedVal, data=total_M_clean, family=binomial(link = logit))
summary(model)##
## Call:
## glm(formula = acceptReject ~ summedVal, family = binomial(link = logit),
## data = total_M_clean)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -3.4096 -0.3046 0.0619 0.4139 3.6912
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 0.49218 0.07588 6.486 8.82e-11 ***
## summedVal 3.20337 0.14653 21.861 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 2777.6 on 2049 degrees of freedom
## Residual deviance: 1144.5 on 2048 degrees of freedom
## AIC: 1148.5
##
## Number of Fisher Scoring iterations: 7xv <- seq(min(total_M_clean$summedVal), max(total_M_clean$summedVal), 0.01)
yv <- predict(model,list(summedVal=xv), type="response")
lines(xv,yv)
#Find the inflection point where there is a 50/50 probability of subject accepting.
p <- 0.5
x <- (log(p/(1-p)) - coef(model)[1]) / coef(model)[2]
x## (Intercept)
## -0.153645GLM STUFF
t <- lm(correct ~ faceVal * houseVal, data=total_M_clean)
anova(t)## Analysis of Variance Table
##
## Response: correct
## Df Sum Sq Mean Sq F value Pr(>F)
## faceVal 1 0.259 0.2594 2.7280 0.09876 .
## houseVal 1 0.189 0.1892 1.9893 0.15857
## faceVal:houseVal 1 7.639 7.6386 80.3298 < 2e-16 ***
## Residuals 2046 194.555 0.0951
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1t <- lm(acceptReject ~ faceVal * houseVal, data=total_M_clean)
anova(t)## Analysis of Variance Table
##
## Response: acceptReject
## Df Sum Sq Mean Sq F value Pr(>F)
## faceVal 1 140.810 140.810 1148.86 < 2.2e-16 ***
## houseVal 1 98.304 98.304 802.06 < 2.2e-16 ***
## faceVal:houseVal 1 6.638 6.638 54.16 2.663e-13 ***
## Residuals 2046 250.767 0.123
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1fit <- lm(correct ~ faceVal + houseVal, data=total_M)
summary(fit) ##
## Call:
## lm(formula = correct ~ faceVal + houseVal, data = total_M)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.91138 0.09702 0.11062 0.12396 0.14697
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.885193 0.006947 127.413 <2e-16 ***
## faceVal 0.020598 0.011878 1.734 0.0831 .
## houseVal 0.013209 0.011729 1.126 0.2602
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.3181 on 2097 degrees of freedom
## Multiple R-squared: 0.002039, Adjusted R-squared: 0.001087
## F-statistic: 2.142 on 2 and 2097 DF, p-value: 0.1177fit <- lm(acceptReject ~ faceVal + houseVal, data=total_M)
summary(fit) ##
## Call:
## lm(formula = acceptReject ~ faceVal + houseVal, data = total_M)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.21035 -0.27325 -0.00131 0.28556 1.08516
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.574028 0.007809 73.51 <2e-16 ***
## faceVal 0.444957 0.013351 33.33 <2e-16 ***
## houseVal 0.367391 0.013184 27.87 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.3575 on 2097 degrees of freedom
## Multiple R-squared: 0.4742, Adjusted R-squared: 0.4737
## F-statistic: 945.7 on 2 and 2097 DF, p-value: < 2.2e-16Linear model if there is a multiplier on FACES
fit <- lm(acceptReject ~ faceVal + houseVal, data=subset(total_M_clean,mult2Face>1 & mult1House==1))
summary(fit) ##
## Call:
## lm(formula = acceptReject ~ faceVal + houseVal, data = subset(total_M_clean,
## mult2Face > 1 & mult1House == 1))
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.10366 -0.17233 -0.00968 0.20768 0.80228
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.58080 0.01426 40.716 <2e-16 ***
## faceVal 0.66388 0.02464 26.942 <2e-16 ***
## houseVal 0.05225 0.02317 2.255 0.0246 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.2916 on 442 degrees of freedom
## Multiple R-squared: 0.6221, Adjusted R-squared: 0.6204
## F-statistic: 363.8 on 2 and 442 DF, p-value: < 2.2e-16Linear model if there is a multiplier on HOUSES
fit <- lm(acceptReject ~ faceVal + houseVal, data=subset(total_M_clean,mult1House>1 & mult2Face ==1))
summary(fit)##
## Call:
## lm(formula = acceptReject ~ faceVal + houseVal, data = subset(total_M_clean,
## mult1House > 1 & mult2Face == 1))
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.91038 -0.20640 0.02105 0.20452 1.04226
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.52463 0.01626 32.263 < 2e-16 ***
## faceVal 0.13668 0.02663 5.132 4.41e-07 ***
## houseVal 0.64875 0.02806 23.124 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.3264 on 417 degrees of freedom
## Multiple R-squared: 0.5687, Adjusted R-squared: 0.5666
## F-statistic: 274.9 on 2 and 417 DF, p-value: < 2.2e-16T-test on flips trials…no effect…
#T-test compare "flips" to non flips
t.test((total_M_clean$decisionTime[total_M_clean$flip == 1]),(total_M_clean$decisionTime[total_M_clean$flip == 0]))##
## Welch Two Sample t-test
##
## data: (total_M_clean$decisionTime[total_M_clean$flip == 1]) and (total_M_clean$decisionTime[total_M_clean$flip == 0])
## t = -1.0262, df = 71.803, p-value = 0.3082
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -0.7010674 0.2245862
## sample estimates:
## mean of x mean of y
## 2.23673 2.47497#T-test to see if if differnce in RT for mult vs. non mult significant?
#T-Test that checks if the means of decision time with NO multipliers and at least ONE multiplier varies
t.test((total_M_clean$decisionTime[total_M_clean$mult2Face == 1 & total_M_clean$mult1House == 1]),(total_M_clean$decisionTime[total_M_clean$mult2Face > 1 | total_M_clean$mult1House >1]))##
## Welch Two Sample t-test
##
## data: (total_M_clean$decisionTime[total_M_clean$mult2Face == 1 & total_M_clean$mult1House == and (total_M_clean$decisionTime[total_M_clean$mult2Face > 1 | total_M_clean$mult1House > 1]) and 1])
## t = -0.13221, df = 2029.7, p-value = 0.8948
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -0.1575432 0.1376436
## sample estimates:
## mean of x mean of y
## 2.462117 2.472067#T-Test that checks if the means of decision time with NO multipliers and TWO multiplier varies
t.test((total_M_clean$decisionTime[total_M_clean$mult2Face == 1 & total_M_clean$mult1House == 1]),(total_M_clean$decisionTime[total_M_clean$mult2Face > 1 & total_M_clean$mult1House >1]))##
## Welch Two Sample t-test
##
## data: (total_M_clean$decisionTime[total_M_clean$mult2Face == 1 & total_M_clean$mult1House == and (total_M_clean$decisionTime[total_M_clean$mult2Face > 1 & total_M_clean$mult1House > 1]) and 1])
## t = -2.1611, df = 161.12, p-value = 0.03216
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -0.66430767 -0.02992467
## sample estimates:
## mean of x mean of y
## 2.462117 2.809233The standard package of plots
fit <- lm(acceptReject ~ faceVal + houseVal, data=total_M_clean)
summary(fit) ##
## Call:
## lm(formula = acceptReject ~ faceVal + houseVal, data = total_M_clean)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.08561 -0.27063 -0.00188 0.28462 1.08184
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.577366 0.007837 73.67 <2e-16 ***
## faceVal 0.443128 0.013375 33.13 <2e-16 ***
## houseVal 0.369897 0.013230 27.96 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.3546 on 2047 degrees of freedom
## Multiple R-squared: 0.4816, Adjusted R-squared: 0.4811
## F-statistic: 950.8 on 2 and 2047 DF, p-value: < 2.2e-16plot(fit)
