W3: Data Visualization

Question 1

What are the 2 methods of scoring scales?

Answer 1

1. Added together for total sum score
2. Average of all items

Question 2

What are the two ways of averaging item scales using rowMeans()

Answer 2

1. Normal, just average
2. Multiple by number of items after averaging

Question 3

What is .SD

Answer 3

• Refers to Subset (S) of Data (D)
• On its own returns all data you’re working on
  E.g unicorn[ , .SD]
• .SDcols tells data.table what columns you want
  db[, StressAVG := rowMeans(.SD, na.rm = TRUE), .SDcols = c(“PSS1”, “PSS2”, “PSS3”, “PSS4)]

Question 4

What function is used to calculate reliability of a scale?

Answer 4

psych::alpha()
* refers to Cronbach’s alpha

Question 5

What should you add to psych::alpha when using reverse scored scales?

Answer 5

• check.keys = TRUE
  E.g psych::alpha( as.data.frame( db
  [, .(PSS1, PSS2r, PSS3r, PSS4)]),
  check.keys = TRUE)

Question 6

What do aesthetics do and what are 4 examples of them in ggplot2?

Answer 6

• Controls how geometrics are displayed
• Size, shape, colour, transparency level

Question 7

What are 4 common geoms_ used for univariate graphs?

Answer 7

geom_histogram( ) , geom_density( ) , geom_dotplot( ), geom_qq( )

Question 8

What argument does geom_qq() need?

Answer 8

scale( predictor ) to z-score data
* z = (x-mean) / SD
geom_abline( intercept = 0, slope = 1): line where all points would fall if normally distributed

Question 9

What function is used to check distribution?

Answer 9

plot( testDistribution() )
E.g plot(testDistribution(db$Stress,
extremevalues = “theoretical”, ev.perc = .005))

Question 10

What 3plots are shown from using plot( testDistribution() )?

Answer 10

Density plot, rug plot, deviates plot

Question 11

What do you need to do when mapping additional categorical variables onto graphs?

Answer 11

Convert variable into a factor
* db [, sex := factor ( sex, levels = c(1,2),
labels = c(“male”, “female))]
* ggplot(db[!is.na(sex)], aes(Stress, colour = sex)) + geom_density()

Question 12

When using geom_histogram, it is more helpful to control what?

Answer 12

Fill colour
e.g ggplot(db[!is.na(sex)], aes(Stress, fill = sex)) +
geom_histogram()

Question 13

What is the argument to have bars side by side when using geom_historgram?

Answer 13

geom_histogram(position = “dodge”)
* bars are stacked by default

Question 14

What are 3 common geoms used for bivariate graphs?

Answer 14

geom_point() scatter plot, geom_line(), geom_bar(stat = “identity”) for values to be actual bar height

Question 15

What is best practice for data visualization?

Answer 15

More data, less ink

Question 16

What are 4 ways to reduce ink and provide more data in graphs?

Answer 16

1. Remove background borders - theme_pubr()
2. Remove axis lines - theme(axis.line = element_blank() )
3. Replace geom_bar with geom_point
4. Using shapes for values - scale_shape_manual(
   name = “Sex”,
   values = c(“male” = 1, “female” = 3))

Question 17

How do you change axes to only go the range of observed data?

Answer 17

Using geom_rangeframe()

Question 18

What are 2 ways to add [interquartile] break points to axis labels?

Answer 18

1. Using quantile()
   * scale_x_continuous(breaks = as.numeric(quantile(db$Stress)))
   * scale_y_continuous(breaks = as.numeric(quantile(db$SE)))
2. Using scale_x/y_discrete
   scale_y_continuous(labels = percent) +
   scale_x_discrete(
   breaks = c(“High SE”, “Low SE”),
   labels = c(“High SE (median)”, “Low SE (median)”))

Question 19

What are the functions used for boxplot with raw data shown?

Answer 19

geom_boxplot() + geom_jitter()

Question 20

What are the 2 ways to provide mean and/or 95% CI on graphs?

Answer 20

1. stat_summary(fun.data = mean_cl_normal)
2. Using prop.test
   LL = prop.test
   (x = sum(sex == “female”, na.rm = TRUE), n = sum(!is.na(sex)), correct = FALSE)$conf.int[1],
   UL = prop.test(
   x = sum(sex == “female”, na.rm = TRUE),
   n = sum(!is.na(sex)), correct =FALSE)$conf.int[2])

Question 21

What should you do before graphing all categorical variables?

Answer 21

Make 1 variable “continuous” by getting their percentages using egltable()

Question 22

What function provides multi-panel plot which is useful for all categorical variables graphing?

Answer 22

facet_grid and/or coord_flip

Question 23

What is the common graph/geom for all continuous variables?

Answer 23

geom_point i.e scatter plot

Question 24

What are 4 things you can add to a graph with all continuous variables?

Answer 24

1. correlation coeff and p-values using
   cor.test(~ SE + Stress, data = db)
2. regression line using
   stat_smooth(method = “lm”)
3. text annotation using
   annotate(“text”, x = max(db$Stress), y = max(db$SE),
   label = “r = -0.65, p < .001”,
   size = 6, hjust = 1, vjust = 1)
4. histograms to margins using
   ggMarginal( x, type = “histogram”)

Question 25

How do you make more space for long axis labels?

Answer 25

ggarrange( ggtitle ( “rotate text”) or (“rotate graph”)

Question 26

What are 3 ways to improve geom_dotplot visualizations?

Answer 26

1. binwidth = .1 to shrink dot size
2. alpha = .2 for dot transparency
3. y = jitter to add noise of scores

Question 27

What are 2 scenarios you would use geom_violin?

Answer 27

1. For large datasets
2. To compare distributions across variables

Question 28

What is a benefit of using rowMeans instead of simply adding all variable scores together?

Answer 28

it imputes the mean for a person with missing data