.Rhistory

#Q5: How much effort and how hard had you to work from not hard at all to very hard
#Q6: Frustration from not frustrated at all to highly frustrated
tlx_average.perQuestionPrep <- matrix(nrow = 6, ncol = 3, byrow = FALSE)
colnames(tlx_average.perQuestionPrep) <-  c("Question","Pen and Paper", "VR")
tlx_average.perQuestion <- as.data.frame(tlx_average.perQuestionPrep)
#setting up the domains of the nasa-tlx as well as the drawing tool
TLX_questions <- c("mental","physical","time","performance","effort","frustration")
drawingtypes <- c("2D","3D")
#saving the average values for the single domains split by drawing task to the dataframe
for(q in TLX_questions){
tlx_average.perQuestion[match(q,TLX_questions),1] <- q
for (d in drawingtypes){
print(mean(tlx_data[tlx_data$DrawingType==d,q]))
tlx_average.perQuestion[match(q,TLX_questions),match(d,drawingtypes)+1] <- mean(tlx_data[tlx_data$DrawingType==d,q])
}
}
tlx_average.perQuestionMelted = melt(tlx_average.perQuestion, id.vars = "Question") #prepare the dataframe for the plot
#TODO:
# y min und y max für confidence intervals and find param for adding these intervals to the plot 95 percent intervalls
# https://r-graph-gallery.com/4-barplot-with-error-bar.html
ggplot(data = tlx_average.perQuestionMelted, mapping = aes(x = factor(Question,
levels = c("mental","physical","time","performance","effort","frustration"))
, y = value, fill = variable)) +
geom_col(position = position_dodge())+
labs(title= "Average Values of RAW-TLX Domains",
subtitle = "by the Drawing Tool",x = "domains",y="average ratings")+
scale_fill_manual("Drawing Tool", values = c("Pen and Paper" = "darkcyan", "VR" = "orange"))
#TODO: use the 95 percent interval and not the standart deviation
ggplot(data = d, mapping = aes(x = Question, y = value, fill = variable)) +
geom_col(position = position_dodge())+geom_errorbar( aes(x=Question, ymin=value-TLX.perQuestion_std_list, ymax=value+TLX.perQuestion_std_list), colour="black",position = position_dodge(width = 0.9))+
scale_fill_manual("Drawing Tool", values = c("Pen and Paper" = "darkcyan", "VR" = "orange"))
ggplot(tlx_data, aes(x = average, y=DrawingType, colour = DrawingType)) +  # Change filling color
geom_boxplot()+ theme_classic()+theme( axis.text.y=element_blank(),
axis.ticks.y=element_blank(), axis.line.y = element_blank(), axis.text.x = element_text(size = 12), axis.title.y =element_blank()
) +labs(x = "average TLX-score")+
scale_color_manual(name   =" Drawingtool",
labels = c("Pen and Paper", "VR"),
values = c("darkcyan", "orange"))
summary(tlx_data[tlx_data$DrawingType=="2D", ]$average)
sd(tlx_data[tlx_data$DrawingType=="2D", ]$average)
summary(tlx_data[tlx_data$DrawingType=="3D", ]$average)
sd(tlx_data[tlx_data$DrawingType=="3D", ]$average)
# base R function for t-test requires a different df format
tlx_data_wide <- tlx_data %>%
select(ID, DrawingType, average) %>%
spread(DrawingType, average)
t.test(tlx_data_wide$`3D`, tlx_data_wide$`2D`, paired=TRUE)
#mental domain
tlx_mental_wide <- tlx_data %>%
select(ID, DrawingType, mental) %>%
spread(DrawingType, mental)
t.test(tlx_mental_wide$`3D`, tlx_mental_wide$`2D`, paired=TRUE)
#physical domain
tlx_physical_wide <- tlx_data %>%
select(ID, DrawingType, physical) %>%
spread(DrawingType, physical)
t.test(tlx_physical_wide$`3D`, tlx_physical_wide$`2D`, paired=TRUE)
#time domain
tlx_time_wide <- tlx_data %>%
select(ID, DrawingType, time) %>%
spread(DrawingType, time)
t.test(tlx_time_wide$`3D`, tlx_time_wide$`2D`, paired=TRUE)
#performance domain - higher value means higher dissatisfaction
tlx_performance_wide <- tlx_data %>%
select(ID, DrawingType, performance) %>%
spread(DrawingType, performance)
t.test(tlx_performance_wide$`3D`, tlx_performance_wide$`2D`, paired=TRUE)
#effort domain
tlx_effort_wide <- tlx_data %>%
select(ID, DrawingType, effort) %>%
spread(DrawingType, effort)# base R function for t-test requires a different df format
t.test(tlx_effort_wide$`3D`, tlx_effort_wide$`2D`, paired=TRUE)
#frustration domain
tlx_frustration_wide <- tlx_data %>%
select(ID, DrawingType, frustration) %>%
spread(DrawingType, frustration)
t.test(tlx_frustration_wide$`3D`, tlx_frustration_wide$`2D`, paired=TRUE)
tlx_2D.prep <- matrix(nrow = participant_count, ncol = 6, byrow = FALSE)
colnames(tlx_2D.prep) <- c("effort","frustration","mental","performance","physical","time")
tlx_2D <- as.data.frame(tlx_2D.prep)
tlx_3D.prep <- matrix(nrow = participant_count, ncol = 6, byrow = FALSE)
colnames(tlx_3D.prep) <- c("effort","frustration","mental","performance","physical","time", "sum", "average")
tlx_3D.prep <- matrix(nrow = participant_count, ncol = 6, byrow = FALSE)
colnames(tlx_3D.prep) <- c("effort","frustration","mental","performance","physical","time")
tlx_3D <- as.data.frame(tlx_3D.prep)
for(id in 1:participant_count){
tlx_2D[id,1] <- as.numeric(questionnaire[id,paste("PnP",1, sep = "_")])
tlx_2D[id,2] <- as.numeric(questionnaire[id,paste("PnP",2, sep = "_")])
tlx_2D[id,3] <- as.numeric(questionnaire[id,paste("PnP",3, sep = "_")])
tlx_2D[id,4] <- as.numeric(questionnaire[id,paste("PnP",4, sep = "_")])
tlx_2D[id,5] <- as.numeric(questionnaire[id,paste("PnP",5, sep = "_")])
tlx_2D[id,6] <- as.numeric(questionnaire[id,paste("PnP",6, sep = "_")])
tlx_3D[id,1] <- as.numeric(questionnaire[id,paste("VR",1, sep = "_")])
tlx_3D[id,2] <- as.numeric(questionnaire[id,paste("VR",2, sep = "_")])
tlx_3D[id,3] <- as.numeric(questionnaire[id,paste("VR",3, sep = "_")])
tlx_3D[id,4] <- as.numeric(questionnaire[id,paste("VR",4, sep = "_")])
tlx_3D[id,5] <- as.numeric(questionnaire[id,paste("VR",5, sep = "_")])
tlx_3D[id,6] <- as.numeric(questionnaire[id,paste("VR",6, sep = "_")])
}
pca_2D <- prcomp(tlx_2D, scale = TRUE)
#reverse the signs
pca_2D$rotation <- -1*pca_2D$rotation
pca_2D$rotation
#display the first six scores
head(pca_2D$x)
#calculate total variance explained by each principal component
pca_2D_explained <- pca_2D$sdev^2 / sum(pca_2D$sdev^2)
#create scree plot
qplot(c(1:6), pca_2D_explained) +
geom_line() +
xlab("Principal Component") +
ylab("Variance Explained") +
ggtitle("Scree Plot") +
ylim(0, 1)
#create scree plot
qplot(c(1:6), pca_2D_explained) +
geom_line() +
xlab("Principal Component") +
ylab("Variance Explained") +
ggtitle("Scree Plot") +
ylim(0, 1)
#Do the same for 3D sktech maps
pca_3D <- prcomp(tlx_3D, scale = TRUE)
#reverse the signs
pca_3D$rotation <- -1*pca_3D$rotation
#reverse the signs of the scores
pca_3D$x <- -1*pca_3D$x
#display the first six scores
head(pca_3D$x)
#look at the pca result
biplot(pca_3D, scale = 0)
#calculate total variance explained by each principal component
pca_3D_explained <- pca_3D$sdev^2 / sum(pca_3D$sdev^2)
#create scree plot
qplot(c(1:6), pca_3D_explained) +
geom_line() +
xlab("Principal Component") +
ylab("Variance Explained") +
ggtitle("Scree Plot") +
ylim(0, 1)
#create scree plot
qplot(c(1:6), pca_3D_explained) +
geom_line() +
xlab("Principal Component") +
ylab("Variance Explained") +
ggtitle("Scree Plot") +
ylim(0, 1)
drawings_2d.import <- read.csv(here("data/Analysis_2D.csv"))# import the csv that contains the analysis data for the 2D drawings
drawings_3d.import <- read.csv(here("data/Analysis_3D.csv")) # import the csv that contains the analysis data for the 3D drawings
duringThesisNotes.import <- read.csv(here("data/DuringThesisNotes.csv"))
# transpose the tables using data.table library
drawings_2d <- transpose(drawings_2d.import)
drawings_3d <- transpose(drawings_3d.import)
duringThesisNotes <- transpose(duringThesisNotes.import)
# and also shift the col and row names for both lists
rownames(drawings_2d) <- colnames(drawings_2d.import)
colnames(drawings_2d) <- rownames(drawings_2d.import)
rownames(drawings_3d) <- colnames(drawings_3d.import)
colnames(drawings_3d) <- rownames(drawings_3d.import)
rownames(duringThesisNotes) <- colnames(duringThesisNotes.import)
colnames(duringThesisNotes) <- rownames(duringThesisNotes.import)
# after the transpose the correct colnames are in the first row.
# Use the first row as names and delete the row for both lists
names(drawings_2d)<-drawings_2d[1,]
drawings_2d<-drawings_2d[-1,]
names(drawings_3d)<-drawings_3d[1,]
drawings_3d<-drawings_3d[-1,]
names(duringThesisNotes)<-duringThesisNotes[1,]
duringThesisNotes<-duringThesisNotes[-1,]
#create vectors for saving the sum values of the analysis with the length of the number of participants
#TODO: replace these overall values with f-score
allSum2d <- vector(length=participant_count)
allSum3d <- vector(length=participant_count)
#create more vectors for saving only the visibility sum and the correctness sum
all_Visibility2D <- vector(length=participant_count)
all_Z_Visibility2D <- vector(length=participant_count) #thinking about only analyzing the z visibility
all_Correct2D <- vector(length=participant_count)
all_Visibility3D <- vector(length=participant_count)
all_Correct3D <- vector(length=participant_count)
for(id in 1:participant_count){
#counter variables for the loop
sum2d <- 0
sum2Dvisible <- 0
sum2Dcorrect <-0
sum3d <- 0
sum3Dvisible <- 0
sum3Dcorrect <-0
# evaluate the object relations
for(i in 4:129){
# check how many are correct in 2D
sum2d <- sum2d + as.numeric(drawings_2d[id,i])
# check how many are correct in 3D
sum3d <- sum3d + as.numeric(drawings_3d[id,i])
# compare only the visibility
if ( substr(colnames(drawings_2d)[i],1, 1) == "V" ){
sum2Dvisible <- sum2Dvisible + as.numeric(drawings_2d[id,i])
sum3Dvisible <- sum3Dvisible + as.numeric(drawings_3d[id,i])
}
# compare only the correctness
#TODO: replace the absolute correctness with relative values depending on the visibility or different
if ( substr(colnames(drawings_2d)[i],1, 1) == "C" ){
sum2Dcorrect <- sum2Dcorrect + as.numeric(drawings_2d[id,i])
sum3Dcorrect <- sum3Dcorrect + as.numeric(drawings_3d[id,i])
}
}
#save the values to the vectors
allSum2d[id]<-sum2d
allSum3d[id]<-sum3d
all_Visibility2D[id]<-sum2Dvisible
all_Visibility3D[id]<-sum3Dvisible
all_Correct2D[id]<-sum2Dcorrect
all_Correct3D[id]<-sum3Dcorrect
}
#the number of relations that were analysed
# there were 7 objects compared among each other and three dimensions resulting in 63 relations
totalNumberOfRelations <- 63
users_data.prep <- matrix(nrow = participant_count*2, ncol = 7, byrow = FALSE)
colnames(users_data.prep) <- c("ID","IPT","DrawingType","order","visibility.score","correctness.score","f.score")
users_data_frame <- as.data.frame(users_data.prep)
#set up a data frame with the visibility, correctness and f-score average for every participant
users_data.prepAverage <- matrix(nrow = participant_count, ncol = 6, byrow = FALSE)
colnames(users_data.prepAverage) <- c("ID","IPT","order","visibility.score","correctness.score","f.score")
users_data_frame.averages <- as.data.frame(users_data.prepAverage)
users_data.prepFscore <- matrix(nrow = participant_count*2, ncol = 5, byrow = FALSE)
colnames(users_data.prepFscore) <- c("ID","DrawingType","precision","recall","f.score")
users_data_frame.fscore <- as.data.frame(users_data.prepFscore)
#writing the values to the dataframe
for(id in 1:participant_count){
#f-score for 2D
precision.2D <- (all_Correct2D[id]/all_Visibility2D[id])
recall.2D <- all_Correct2D[id]/totalNumberOfRelations
fscore.2D <- 2*((precision.2D*recall.2D)/(precision.2D+recall.2D))
#f-score for 3D
precision.3D <- (all_Correct3D[id]/all_Visibility3D[id])
recall.3D <- all_Correct3D[id]/totalNumberOfRelations
fscore.3D <- 2*((precision.3D*recall.3D)/(precision.3D+recall.3D))
#write data to the dataframes
users_data_frame[id,1] <- as.character(id)
users_data_frame[id,2] <- ipt_results[id]
users_data_frame[id,3] <- "2D"
users_data_frame[id,4] <- if(id %% 2 == 0){"3Dfirst"}else{"2Dfirst"}
users_data_frame[id,5] <- all_Visibility2D[id]
users_data_frame[id,6] <- precision.2D*100
users_data_frame[id,7] <- fscore.2D
users_data_frame[(id+participant_count),1] <- as.character(id)
users_data_frame[(id+participant_count),2] <- ipt_results[id]
users_data_frame[(id+participant_count),3] <- "3D"
users_data_frame[(id+participant_count),4] <- if(id %% 2 == 0){"3Dfirst"}else{"2Dfirst"}
users_data_frame[(id+participant_count),5] <- all_Visibility3D[id]
users_data_frame[(id+participant_count),6] <- precision.3D*100
users_data_frame[(id+participant_count),7] <- fscore.3D
users_data_frame.averages[id,1] <- as.character(id)
users_data_frame.averages[id,2] <- ipt_results[id]
users_data_frame.averages[id,3] <- if(id %% 2 == 0){"3Dfirst"}else{"2Dfirst"}
users_data_frame.averages[id,4] <- mean(c(all_Visibility2D[id],all_Visibility3D[id]))
users_data_frame.averages[id,5] <- mean(c((precision.2D*100),(precision.3D*100)))
users_data_frame.averages[id,6] <- mean(c(fscore.2D,fscore.3D))
users_data_frame.fscore[id,1] <- as.character(id)
users_data_frame.fscore[id,2] <- "2D"
users_data_frame.fscore[id,3] <- precision.2D
users_data_frame.fscore[id,4] <- recall.2D
users_data_frame.fscore[id,5] <- fscore.2D
users_data_frame.fscore[(id+participant_count),1] <- as.character(id)
users_data_frame.fscore[(id+participant_count),2] <- "3D"
users_data_frame.fscore[(id+participant_count),3] <- precision.3D
users_data_frame.fscore[(id+participant_count),4] <- recall.3D
users_data_frame.fscore[(id+participant_count),5] <- fscore.3D
}
#mutate drawing type, Id and order into factors
users_data_frame <- users_data_frame %>%
mutate(DrawingType = as.factor(DrawingType),
ID = as.factor(ID),
order = as.factor(order))
users_data_frame.averages <- users_data_frame.averages %>%
mutate(ID = as.factor(ID),
order = as.factor(order))
users_data_frame.fscore <- users_data_frame.fscore %>%
mutate(ID = as.factor(ID))
ggplot(users_data_frame, aes(IPT, visibility.score)) +
geom_point()+
geom_smooth(method = lm,
se     = FALSE,
size   = .5,
alpha  = .8)+ # to add regression line
labs(title    = "Visibility-Score Depending on the IPT")
#setting up the model
IPTmodel.visibility <- lm(visibility.score ~ IPT, data=users_data_frame.averages)
summary(IPTmodel.visibility)
#residual plot
plot(fitted(IPTmodel.visibility),residuals(IPTmodel.visibility))
#checking the correctness-score dependency from the IPT
ggplot(users_data_frame, aes(IPT, correctness.score)) +
geom_point()+
geom_smooth(method = lm,
se     = FALSE,
size   = .5,
alpha  = .8)+ # to add regression line
labs(title    = "Correctness-Score Depending on the IPT")
IPTmodel.correctness <- lm(correctness.score ~ IPT, data=users_data_frame.averages)
summary(IPTmodel.correctness)
#residual plot
plot(fitted(IPTmodel.correctness),residuals(IPTmodel.correctness))
#checking the f-score dependency from the IPT
ggplot(users_data_frame, aes(IPT, f.score)) +
geom_point()+
geom_smooth(method = lm,
se     = FALSE,
size   = .5,
alpha  = .8)+ # to add regression line
labs(title    = "F-Score depending on the IPT")
IPTmodel.fscore <- lm(f.score ~ IPT, data=users_data_frame.averages)
summary(IPTmodel.fscore)
#residual plot
plot(fitted(IPTmodel.fscore),residuals(IPTmodel.fscore))
ggplot(users_data_frame, aes(order, visibility.score)) +
geom_point()+
geom_smooth(method = lm,
se     = FALSE,
size   = .5,
alpha  = .8)+ # to add regression line
labs(title    = "Visibility-Score Depending on the order of drawings")
#setting up the model
ordermodel.visibility <- lm(visibility.score ~ order, data=users_data_frame.averages)
summary(ordermodel.visibility)
#residual plot
plot(fitted(ordermodel.visibility),residuals(ordermodel.visibility))
#checking the correctness-score dependency from the IPT
ggplot(users_data_frame, aes(order, correctness.score)) +
geom_point()+
geom_smooth(method = lm,
se     = FALSE,
size   = .5,
alpha  = .8)+ # to add regression line
labs(title    = "Correctness-Score Depending on the order of drawings")
ordermodel.correctness <- lm(correctness.score ~ order, data=users_data_frame.averages)
summary(ordermodel.correctness)
#residual plot
plot(fitted(ordermodel.correctness),residuals(ordermodel.correctness))
#checking the f-score dependency from the IPT
ggplot(users_data_frame, aes(order, f.score)) +
geom_point()+
geom_smooth(method = lm,
se     = FALSE,
size   = .5,
alpha  = .8)+ # to add regression line
labs(title    = "F-Score depending on the order of drawings")
ordermodel.fscore <- lm(f.score ~ order, data=users_data_frame.averages)
summary(ordermodel.fscore)
#residual plot
plot(fitted(ordermodel.fscore),residuals(ordermodel.fscore))
ggplot(users_data_frame, aes(DrawingType, visibility.score)) +
geom_point()+
geom_smooth(method = lm,
se     = FALSE,
size   = .5,
alpha  = .8)+ # to add regression line
labs(title    = "Visibility-Score depending on the drawing type",
subtitle = "all 3D scores are layerd")
#set up data frames containing only 2D or 3D data
users_data_frame.2D <-users_data_frame[users_data_frame$DrawingType=="2D",]
users_data_frame.3D <-users_data_frame[users_data_frame$DrawingType=="3D",]
#summary of 2d sketch maps' visibility-score
summary(users_data_frame.2D$visibility.score)
boxplot(users_data_frame.2D$visibility.score)
#every drawing in 3D has the maximum of 63 points
summary(users_data_frame.3D$visibility.score)
#disable scientific notation to easier compare very low numbers
options(scipen = 999)
#setting up the linear mixed effects model for the visible results only
visibility.model = lmer(visibility.score ~ IPT + order + (1|ID) + DrawingType, data=users_data_frame, REML = FALSE)
summary(visibility.model)
#show that the model is singular
performance::icc(visibility.model)#true
lme4::isSingular(visibility.model)#true
performance::check_singularity(visibility.model)#true
# base R function for t-test requires a different df format
users_data_wide <- users_data_frame %>%
select(ID, IPT, order, DrawingType, visibility.score) %>%
spread(DrawingType, visibility.score)
t.test(users_data_wide$`3D`, users_data_wide$`2D`, paired=TRUE)
#setting up the null model where the drawing type is omitted
visibility.null = lmer(visibility.score ~ IPT + order + (1|ID), data=users_data_frame, REML = FALSE)
#anova for significance
anova(visibility.null,visibility.model)
#residual plot
plot(fitted(visibility.model), resid(visibility.model, type = "pearson"))+
abline(0,0, col="red")
#setting up the null model where the drawing type is omitted
visibility.null = lmer(visibility.score ~ IPT + order + (1|ID), data=users_data_frame, REML = FALSE)
#anova for significance
anova(visibility.null,visibility.model)
#residual plot
plot(fitted(visibility.model), resid(visibility.model, type = "pearson"))+
abline(0,0, col="red")
visibility.modelSimple = lmer(visibility.score ~ (1|ID) + DrawingType, data=users_data_frame, REML = FALSE)
#this is not singular but the null model is
visibility.nullSimple =  lmer(visibility.score ~ (1|ID), data=users_data_frame, REML = FALSE)
anova(visibility.nullSimple,visibility.modelSimple)
#TODO: use a relative value or something else for the correctness
#have a look at the data with the following plots
ggplot(users_data_frame, aes(DrawingType, correctness.score, colour = order)) +
geom_point()+
geom_smooth(method = lm,
se     = FALSE,
size   = .5,
alpha  = .8)+ # to add regression line
labs(title    = "Correctness-Score depending on drawing type",
subtitle = "and in color the different order of drawings")
ggplot(data = users_data_frame,
aes(x   = IPT,
y   = correctness.score,
col = as.factor(DrawingType)))+
geom_point(size     = 1.5,
alpha    = .7,
position = "jitter")+
geom_smooth(method   = lm,
se       = T,
size     = 1.5,
linetype = 1,
alpha    = .3)+
theme_minimal()+
labs(title    = "Correlation of Relative Correctness-Score and IPT-result")+
scale_color_manual(name   =" Drawingtool",
labels = c("Pen and Paper", "VR"),
values = c("darkcyan", "orange"))
#setting up the linear mixed effects model for the correctness score analysis
correctness.model = lmer(correctness.score ~ IPT + order + (1|ID) + DrawingType, data=users_data_frame, REML = FALSE)
#setting up the null model and check the significants of the model with the anova
correctness.null = lmer(correctness.score ~ IPT + order + (1|ID) , data=users_data_frame, REML = FALSE)
correctness.nullWithoutIPT = lmer(correctness.score ~ order + (1|ID) + DrawingType , data=users_data_frame, REML = FALSE)
correctness.nullWithoutOrder = lmer(correctness.score ~ IPT + (1|ID) + DrawingType , data=users_data_frame, REML = FALSE)
correctness.modelOrderOnly = lmer(correctness.score ~ order + (1|ID), data=users_data_frame, REML = FALSE)
correctness.nullOrderOnly = lmer(correctness.score ~  (1|ID), data=users_data_frame, REML = FALSE)
anova(correctness.nullOrderOnly, correctness.modelOrderOnly)
anova(correctness.null,correctness.model) #checking the significants
anova(correctness.nullWithoutIPT,correctness.model) #checking the significants
anova(correctness.nullWithoutOrder,correctness.model) #checking the significants
#residual plot
plot(fitted(correctness.model), resid(correctness.model, type = "pearson"))+
abline(0,0, col="red")
#get an overview
summary(users_data_frame$f.score)
summary(users_data_frame.2D$f.score)
summary(users_data_frame.3D$f.score)
#variance of f-scores of 2D and 3D drawings seperated
var(users_data_frame.2D$f.score)
var(users_data_frame.3D$f.score)
ggplot(users_data_frame, aes(x = f.score, y=DrawingType, colour = DrawingType)) +  # Change filling color
geom_boxplot()+ theme_classic()+theme( axis.text.y=element_blank(),
axis.ticks.y=element_blank(), axis.line.y = element_blank(), axis.text.x = element_text(size = 12), axis.title.y =element_blank()
) +labs(x = "F-Score")+
scale_color_manual(name   =" Drawingtool",
labels = c("Pen and Paper", "VR"),
values = c("darkcyan", "orange"))
#have a look at precision and recall
ggplot(users_data_frame.fscore, aes(precision, recall, colour = DrawingType)) +
geom_point()+
geom_smooth(method = lm,
se     = FALSE,
size   = .5,
alpha  = .8)+ # to add regression line
labs(title    = "Precision and Recall",
subtitle = "colored are the different drawing tools")
ggplot(users_data_frame, aes(IPT, f.score, colour = DrawingType)) +
geom_point()+
geom_smooth(method = lm,
se     = FALSE,
size   = .5,
alpha  = .8)+ # to add regression line
labs(title    = "Precision and Recall",
subtitle = "colored are the different drawing tools")
#setting up the linear mixed effects model
fscore.model = lmer(f.score ~ IPT + order + DrawingType + (1|ID), data=users_data_frame, REML = FALSE)
summary(fscore.model)
#setting up the null model that does not contain the Drawing Type
fscore.null = lmer(f.score ~ IPT + order + (1|ID), data=users_data_frame, REML = FALSE)
#setting up the null model where the ipt is omitted
fscore.nullIPT = lmer(f.score ~ order + DrawingType + (1|ID), data=users_data_frame, REML = FALSE)
#setting up the null model where the order of drawing is omitted
fscore.nullOrder = lmer(f.score ~ IPT + DrawingType + (1|ID), data=users_data_frame, REML = FALSE)
#anova for checking the relevance of the different values
anova(fscore.null,fscore.model) # checking the relevance of the drawing type
anova(fscore.nullIPT,fscore.model) # checking the relevance of the IPT
anova(fscore.nullOrder,fscore.model) # checking the relevance of the order of the drawing
ggplot(data = users_data_frame,
aes(x   = IPT,
y   = f.score,
col = as.factor(DrawingType)))+
geom_point(size     = 1.5,
alpha    = .7,
position = "jitter")+
geom_smooth(method   = lm,
se       = T,
size     = 1.5,
linetype = 1,
alpha    = .3)+
theme_minimal()+
labs(title    = "Correlation of F-Score and IPT-result for the 2 Drawingtools")+
scale_color_manual(name   =" Drawingtool",
labels = c("Pen and Paper", "VR"),
values = c("darkcyan", "orange"))
#residual plot
plot(fitted(fscore.model), resid(fscore.model, type = "pearson"))+
abline(0,0, col="red")
#visible.plottest = lmer(score ~ IPT + (1|ID) + (1|DrawingType), data=users_data_frame.visible)
#estimates(visible.plottest)
#visualize(visible.plottest, plot="all",formula = NULL)
#TODO: is there a way to compare these two?
#TODO: create graphics
users_data_frame.2D
count(users_data_frame.2D$visibility.score == 63)
users_data_frame.2D$visibility.score == 63