stefan-mueller
commented
7 years ago Some general remarks first.
- When you require help, please make your code readable by using line breaks and following the R style guide.
- Provide a minimal worked example - several of your options and commands are not listed.
- Upload the correct file (you uploaded an xlsx file, but load a csv file!).
- Indicate the packages to be loaded.
No offense taken, but if you require help, make sure that we don't have to wrangle your code for 15 minutes.
Why didn't your code work? Basically, because ordering and relabeling needs to be applied to a data frame. Please see my solution below (and some comments).
library(survey)
library(dotwhisker)
library(dplyr)
library(tidyverse)
##Creating the initial dataset (please upload correct data!)
table3 <- read.csv("ABDIII_EXLAPA.csv")
##Subsetting the data to include the columns necessary for analysis
# NOTE: not a good way to subset data -- look at select and filter from dplyr
newdata <- table3[c(1:4,6,8,9,13:31,50:53,61,63:66,70:75,77:79)]
##Removing non-responses
mydata <- na.omit(newdata)
##Model 1d: Protest with survey glm package and binary internet use, org.membership measures, motivations, grievance measures and all other variables
mydata$INCOME_QUINT_2 <- as.factor(mydata$INCOME_QUINT_1)
mydata$EDUC_BAS_2 <- as.factor(mydata$EDUC_BAS)
svydata1d <- svydesign(id = ~PSU, strata = ~STRATA_2,
weights = ~WT, data = mydata)
summary(svydata1d)
options(survey.lonely.psu="adjust")
## Run models
mysvylogit1d <- svyglm(PROTEST ~ INTERNET_1_BIN +
FORM_ORG_BIN + # not in model 1f
POLIT_KNOW +
POLIT_INT +
ABS_GRIEVE +
REL_GRIEVE +
INCOME_QUINT_2 +
QURAN +
GENDER +
SINGLE +
EDUC_BAS_2 +
STUDENT +
EMPLOYED +
AGE,
design = svydata1d,
family=quasibinomial(link=logit))
summary(mysvylogit1d)
mydata$ORG_TYPE_21 <- as.factor(mydata$ORG_TYPE_2)
svydata1f <- svydesign(id= ~PSU, strata= ~STRATA_2, weights=~WT, data=mydata)
options(survey.lonely.psu="adjust")
mysvylogit1f <- svyglm(PROTEST ~ INTERNET_1_BIN +
POLIT_KNOW +
POLIT_INT +
ABS_GRIEVE +
REL_GRIEVE +
INCOME_QUINT_2 +
QURAN +
GENDER +
SINGLE +
EDUC_BAS_2 +
STUDENT +
EMPLOYED +
AGE +
ORG_TYPE_21, # not in model 1e
design = svydata1f,
family = quasibinomial(link=logit))
summary(mysvylogit1f)
# Disply a basic dwplot
dwplot(list(mysvylogit1d, mysvylogit1f))
# works correctly
order_vars_original <- c("INTERNET_1_BIN",
"FORM_ORG_BIN",
"ORG_TYPE_211",
"ORG_TYPE_212",
"POLIT_KNOW",
"POLIT_INT",
"ABS_GRIEVE",
"REL_GRIEVE",
"INCOME_QUINT_22",
"INCOME_QUINT_23",
"INCOME_QUINT_24",
"INCOME_QUINT_25",
"QURAN",
"GENDER",
"SINGLE",
"EDUC_BAS_22",
"EDUC_BAS_23",
"STUDENT",
"EMPLOYED",
"AGE")
data_plot <- rbind(broom::tidy(mysvylogit1d) %>% mutate(model = "Model 1"),
broom::tidy(mysvylogit1f) %>% mutate(model = "Model 2")) %>%
filter(term != "(Intercept)") %>%
mutate(term = factor(term, levels = order_vars_original)) %>%
relabel_predictors(c(INTERNET_1_BIN = "Internet use",
FORM_ORG_BIN = "Organizational membership",
POLIT_KNOW = "Political knowledge",
POLIT_INT = "Political interest",
ABS_GRIEVE = "Absolute grievance",
REL_GRIEVE = "Relative grievance",
INCOME_QUINT_22 = "Income quintile (21-40)",
INCOME_QUINT_23 = "Income quintile (41-60)",
INCOME_QUINT_24 = "Income quintile (61-80)",
INCOME_QUINT_25 = "Income quintile (81-100)",
QURAN = "Religiosity",
GENDER = "Gender",
SINGLE = "Marital status",
EDUC_BAS_22 = "High School Educated",
EDUC_BAS_23 = "College Educated",
STUDENT = "Student",
EMPLOYED = "Employment status",
AGE = "Age",
ORG_TYPE_211 = "Expressive org. membership",
ORG_TYPE_212 = "Instrumental org. membership"))
## Only use each option once (e.g. theme) and improve structure
dwplot(data_plot, dodge_size = 0.8) +
theme_bw() +
ggtitle("Determinants of individual protest participation") +
xlab("Coefficient Estimate") +
ylab("") +
xlim(-1,2) +
geom_vline(xintercept = 0, colour="grey60", linetype=2) +
theme(legend.position = "none") +
annotate("text", x = 1.5, y = 10, size = 3, hjust = 0, label = "R^2 == 0.169", parse = TRUE) + annotate("text", x = 1.5, y = 9, size = 3, hjust = 0, label = "N = 6663") +
theme(axis.text.y = element_text(size = 11),
axis.text.x = element_text(size = 12),
axis.title=element_text(size=14),
plot.title = element_text(face = "bold", size = 14, hjust = 0.5),
legend.position = "none",
panel.grid.minor.y = element_blank(),
panel.grid.major.y=element_blank(),
panel.grid.minor.x = element_blank(),
panel.grid.major.x=element_blank())
Edit: I see that there is still something wrong with the ordering/renaming, but this code seems to work. So go from there and check the parts from order_vars_original onwards. If there are serious issues you cannot solve, please let me know.
dwplot(list(mysvylogit1d, mysvylogit1f))
fsolt
ABDIII_EXLAPA.xlsx I have the attached data and am trying to create a dot and whisker plot using the following lines of code:
Creating the initial dataset
table3 <- read.csv("ABDIII_EXLAPA.csv")
Subsetting the data to include the columns necessary for analysis
newdata <- table3[c(1:4,6,8,9,13:31,50:53,61,63:66,70:75,77:79)]
Removing non-responses
mydata <- na.omit(newdata)
Model 1d: Protest with survey glm package and binary internet use, org.membership measures, motivations, grievance measures and all other variables
mydata$INCOME_QUINT_2 <- as.factor(mydata$INCOME_QUINT_1) mydata$EDUC_BAS_2 <- as.factor(mydata$EDUC_BAS)
svydata1d <- svydesign(id= ~PSU, strata= ~STRATA_2, weights=~WT, data=mydata)
summary(svydata1d)
options(survey.lonely.psu="adjust")
mysvylogit1d <- svyglm(PROTEST ~ INTERNET_1_BIN + FORM_ORG_BIN + POLIT_KNOW + POLIT_INT + ABS_GRIEVE + REL_GRIEVE + INCOME_QUINT_2 + QURAN + GENDER + SINGLE + EDUC_BAS_2 + STUDENT + EMPLOYED + AGE, design = svydata1d, family=quasibinomial(link=logit))
summary(mysvylogit1d)
Model 1f: Protest with survey glm package with all variables and organizational type
mydata$ORG_TYPE_21 <- as.factor(mydata$ORG_TYPE_2)
svydata1f <- svydesign(id= ~PSU, strata= ~STRATA_2, weights=~WT, data=mydata)
summary(svydata1f)
options(survey.lonely.psu="adjust")
mysvylogit1f <- svyglm(PROTEST ~ INTERNET_1_BIN + ORG_TYPE_21+ POLIT_KNOW + POLIT_INT + ABS_GRIEVE + REL_GRIEVE + INCOME_QUINT_2 + QURAN + GENDER + SINGLE + EDUC_BAS_2 + STUDENT + EMPLOYED + AGE, design = svydata1f, family=quasibinomial(link=logit))
summary(mysvylogit1f)
coefplot1 <- dwplot(list(mysvylogit1d, mysvylogit1f), dodge_size = 0.8, order_vars = c("INTERNET_1_BIN", "FORM_ORG_BIN", "ORG_TYPE_211", "ORG_TYPE_212", "POLIT_KNOW", "POLIT_INT", "ABS_GRIEVE", "REL_GRIEVE", "INCOME_QUINT_22", "INCOME_QUINT_23", "INCOME_QUINT_24", "INCOME_QUINT_25", "QURAN", "GENDER", "SINGLE", "EDUC_BAS_22", "EDUC_BAS_23", "STUDENT", "EMPLOYED", "AGE")) %>% relabel_predictors(c(INTERNET_1_BIN = "Internet use", FORM_ORG_BIN = "Organizational membership", POLIT_KNOW = "Political knowledge", POLIT_INT = "Political interest", ABS_GRIEVE = "Absolute grievance", REL_GRIEVE = "Relative grievance", INCOME_QUINT_22 = "Income quintile (21-40)", INCOME_QUINT_23 = "Income quintile (41-60)", INCOME_QUINT_24 = "Income quintile (61-80)", INCOME_QUINT_25 = "Income quintile (81-100)", QURAN = "Religiosity", GENDER = "Gender", SINGLE = "Marital status", EDUC_BAS_22 = "High School Educated", EDUC_BAS_23 = "College Educated", STUDENT = "Student", EMPLOYED = "Employment status", AGE = "Age", ORG_TYPE_211 = "Expressive org. membership", ORG_TYPE_212 = "Instrumental org. membership")) coefplot1 <- coefplot1 + theme_bw() + theme(panel.grid.minor.y = element_blank(), panel.grid.major.y=element_blank(), panel.grid.minor.x = element_blank(), panel.grid.major.x=element_blank()) + xlab("Coefficient Estimate") + ylab("") + xlim(-1,2) + geom_vline(xintercept = 0, colour="grey60", linetype=2) + theme(legend.position = "none") + annotate("text", x = 1.5, y = 10, size = 3, hjust = 0, label = "R^2 == 0.169", parse = TRUE) + annotate("text", x = 1.5, y = 9, size = 3, hjust = 0, label = "N = 6663") + theme(axis.text.y = element_text(family="CMUSansSerif", size = 11), axis.text.x = element_text(family="CMUSansSerif", size = 12), axis.title=element_text(family = "CMUSansSerif", size=14)) + ggtitle("Determinants of individual protest participation") + theme(plot.title = element_text(family = "CMUSansSerif", face = "bold", size = 14, hjust = 0.5))
However, the final plot does not display all estimates for model 2. Is there a way to fix this? Also even if I remove the order_vars specification, the dodge command doesn't create the spacing I need. Is there a way to rectify this issue as well?