Psych 251 PS5: Visualization
Mike Frank
2018
Intro
library(tidyverse)## ── Attaching packages ────────────────## ✔ ggplot2 3.0.0 ✔ purrr 0.2.5
## ✔ tibble 1.4.2 ✔ dplyr 0.7.6
## ✔ tidyr 0.8.1 ✔ stringr 1.3.1
## ✔ readr 1.1.1 ✔ forcats 0.3.0## ── Conflicts ─────────────────────────
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()library(ggthemes)
theme_set(theme_few())
sem <- function(x) {sd(x, na.rm=TRUE) / sqrt(sum(!is.na((x))))}
ci <- function(x) {sem(x) * 1.96} # reasonable approximation This is problem set #5, in which we hope you will practice the visualization package ggplot2, as well as hone your knowledge of the packages tidyr and dplyr. You’ll look at two different datasets here.
First, data on children’s looking at social targets from Frank, Vul, Saxe (2011, Infancy).
Second, data from Sklar et al. (2012) on the unconscious processing of arithmetic stimuli.
In both of these cases, the goal is to poke around the data and make some plots to reveal the structure of the dataset.
Part 1
This part is a warmup, it should be relatively straightforward ggplot2 practice.
Load data from Frank, Vul, Saxe (2011, Infancy), a study in which we measured infants’ looking to hands in moving scenes. There were infants from 3 months all the way to about two years, and there were two movie conditions (Faces_Medium, in which kids played on a white background, and Faces_Plus, in which the backgrounds were more complex and the people in the videos were both kids and adults). An eye-tracker measured children’s attention to faces. This version of the dataset only gives two conditions and only shows the amount of looking at hands (other variables were measured as well).
fvs <- read_csv("data/FVS2011-hands.csv")## Parsed with column specification:
## cols(
## subid = col_integer(),
## age = col_double(),
## condition = col_character(),
## hand.look = col_double()
## )First, use ggplot to plot a histogram of the ages of children in the study. NOTE: this is a repeated measures design, so you can’t just take a histogram of every measurement.
fvs_wide = fvs %>%
spread(condition, hand.look)
ggplot(fvs_wide, aes(age)) +
geom_histogram(bins=30) +
xlab("Age Months") +
ylab("Frequency")
Second, make a scatter plot showing the difference in hand looking by age and condition. Add appropriate smoothing lines. Take the time to fix the axis labels and make the plot look nice.
ggplot(fvs, aes(x=age, y=hand.look, color=condition)) + geom_point() +
geom_smooth(method=lm,
se=TRUE) +
xlab("Age (months)") +
ylab("Hand Looking") +
xlim(min(4),max(24)) +
scale_x_continuous(breaks = round(seq(min(fvs$age), max(fvs$age + 1), by = 5))) +
scale_color_hue(labels= c( "Medium Faces", "Plus Faces")) +
scale_y_continuous(breaks = round(seq(min(fvs$hand.look), max(fvs$hand.look + .5), by = .05),2)) +
ggtitle("Hand Looking by Age and Condition") +
theme(plot.title=element_text(hjust=.5))## Scale for 'x' is already present. Adding another scale for 'x', which
## will replace the existing scale.
What do you conclude from this pattern of data?
There is a significant interaction between age and condition. Older participants (~ age 13 months - 28 months) in the Plus Faces condition showed significantly more hand looking compared to older children in the Medium Faces condition. There was not a significant difference in the amount of hand looking in young children (~ age 3 months - ~12 months) in the Medium Faces compared to the Plus Faces condition. There is also a main effect of age, such that older children tend to increase in the amount of hand looking regardless of condition.
What statistical analyses would you perform here to quantify these differences?
A regression model with an interaction term between age and condition. Age should also be a predictor variable in the regression model because there is a main effect of age.
Part 2
Sklar et al. (2012) claim evidence for unconscious arithmetic processing - they prime participants with arithmetic problems and claim that the authors are faster to repeat the answers. We’re going to do a reanalysis of their Experiment 6, which is the primary piece of evidence for that claim. The data are generously shared by Asael Sklar. (You may recall these data from the tidyverse tutorial earlier in the quarter).
Data Prep
First read in two data files and subject info. A and B refer to different trial order counterbalances.
subinfo <- read_csv("data/sklar_expt6_subinfo_corrected.csv")## Parsed with column specification:
## cols(
## subid = col_integer(),
## presentation.time = col_integer(),
## subjective.test = col_integer(),
## objective.test = col_double()
## )d_a <- read_csv("data/sklar_expt6a_corrected.csv")## Parsed with column specification:
## cols(
## .default = col_integer(),
## prime = col_character(),
## congruent = col_character(),
## operand = col_character()
## )## See spec(...) for full column specifications.d_b <- read_csv("data/sklar_expt6b_corrected.csv")## Parsed with column specification:
## cols(
## .default = col_integer(),
## prime = col_character(),
## congruent = col_character(),
## operand = col_character()
## )
## See spec(...) for full column specifications.gather these datasets into long (“tidy data”) form. If you need to review tidying, here’s the link to R4DS (bookmark it!). Remember that you can use select_helpers to help in your gathering.
Once you’ve tidied, bind all the data together. Check out bind_rows.
The resulting tidy dataset should look like this:
prime prime.result target congruent operand distance counterbalance subid rt
<chr> <int> <int> <chr> <chr> <int> <int> <dbl> <int>
1 =1+2+5 8 9 no A -1 1 1 597
2 =1+3+5 9 11 no A -2 1 1 699
3 =1+4+3 8 12 no A -4 1 1 700
4 =1+6+3 10 12 no A -2 1 1 628
5 =1+9+2 12 11 no A 1 1 1 768
6 =1+9+3 13 12 no A 1 1 1 595d_a_tidy <- d_a %>%
gather ("subid", "rt",
"1":"21")
d_b_tidy <- d_b %>%
gather("subid", "rt",
"22":"42")Merge these with subject info. You will need to look into merge and its relatives, left_ and right_join. Call this dataframe d, by convention.
d_tidy = merge (d_a_tidy, d_b_tidy, all.x=T, all.y=T)
d_tidy$subid=as.numeric (d_tidy$subid)
subinfo$subid=as.numeric (subinfo$subid)
d = right_join (d_tidy, subinfo, by = "subid")Clean up the factor structure (just to make life easier). No need to, but if you want, you can make this more tidyverse-ish.
d$presentation.time <- factor(d$presentation.time)
levels(d$operand) <- c("addition","subtraction")Data Analysis Preliminaries
Examine the basic properties of the dataset. First, show a histogram of reaction times.
# N/A values for n = 237 participants' RT in the dataframe. The histogram is removing the missing values.
ggplot(d, aes(rt)) +
geom_histogram(bins=50) +
xlab("Reaction Time") +
ylab("Frequency")## Warning: Removed 237 rows containing non-finite values (stat_bin).
Challenge question: what is the sample rate of the input device they are using to gather RTs?
The sample rate is around 30 to 35 units. If you order reaction time and manually look at the data, the reaction times jump around 30 to 35 units per block.
Sklar et al. did two manipulation checks. Subjective - asking participants whether they saw the primes - and objective - asking them to report the parity of the primes (even or odd) to find out if they could actually read the primes when they tried. Examine both the unconscious and conscious manipulation checks. What do you see? Are they related to one another?
Yes, the unconscious and conscious manipulation checks are related to one another (r = 0.58, p < 0.05). This indicates that both manipulation checks were relatively effective.
cor.test(d$subjective.test,d$objective.test)##
## Pearson's product-moment correlation
##
## data: d$subjective.test and d$objective.test
## t = 57.052, df = 6466, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.5622115 0.5946395
## sample estimates:
## cor
## 0.5786542There was a significant (medium-sized) correlation between the unconscious (subjective) and conscious (objective) manipulation checks.
In Experiments 6, 7, and 9, we used the binomial distribution to determine whether each participant performed better than chance on the objective block and excluded from analyses all those participants who did (21, 30, and 7 participants in Experiments 6, 7, and 9, respectively). Note that, although the number of excluded participants may seem high, they fall within the normal range of long-duration CFS priming, in which suc- cessful suppression is strongly affected by individual differences (38). We additionally excluded participants who reported any subjective awareness of the primes (four, five, and three participants in Experiments 6, 7, and 9, respectively).
OK, let’s turn back to the measure and implement Sklar et al.’s exclusion criterion. You need to have said you couldn’t see (subjective test) and also be not significantly above chance on the objective test (< .6 correct). Call your new data frame ds.
ds = d %>%
filter(subjective.test==0,
objective.test <0.6)Replicating Sklar et al.’s analysis
Sklar et al. show a plot of a “facilitation effect” - the amount faster you are for prime-congruent naming compared with prime-incongruent naming. They then show plot this difference score for the subtraction condition and for the two prime times they tested. Try to reproduce this analysis.
HINT: first take averages within subjects, then compute your error bars across participants, using the ci function (defined above). Sklar et al. use SEM (and do it incorectly, actually), but CI is more useful for “inference by eye” as discussed in class.
HINT 2: remember that in class, we reviewed the common need to group_by and summarise twice, the first time to get means for each subject, the second time to compute statistics across subjects.
HINT 3: The final summary dataset should have 4 rows and 5 columns (2 columns for the two conditions and 3 columns for the outcome: reaction time, ci, and n).
ds_average = ds %>%
filter (operand== "S") %>%
group_by (subid, congruent, presentation.time) %>%
summarize (rt_average=mean(rt,na.rm=TRUE)) %>%
group_by (congruent, presentation.time) %>%
summarize (rt_average_all = mean(rt_average, na.rm=TRUE), ci=ci(rt_average),
n=n())Now plot this summary, giving more or less the bar plot that Sklar et al. gave (though I would keep operation as a variable here. Make sure you get some error bars on there (e.g. geom_errorbar or geom_linerange).
ds_difference_score_average = ds %>%
filter(operand== "S") %>%
group_by(subid, congruent, presentation.time) %>%
summarize(rt=mean(rt, na.rm=TRUE)) %>%
spread(congruent,rt) %>%
mutate(diff=no-yes) %>%
group_by(presentation.time) %>%
summarize(diff_all=mean(diff),
ci=ci(diff),
n=n())
ggplot(ds_difference_score_average, aes(x=presentation.time, y=diff_all)) +
geom_bar(position="dodge", stat="identity") +
geom_errorbar(aes(ymin=diff_all - ci, ymax = diff_all +ci), width=.2,
position=position_dodge(0.9)) +
ggtitle("Sklar Reproduction") +
theme(plot.title=element_text(hjust=.5)) +
xlab("Facilitation Time") +
ylab("Presentation Time")
What do you see here? How close is it to what Sklar et al. report? How do you interpret these data?
The error bars are much larger than what Sklar et al. reported. It is highly likely that Sklar and colleagues calculated their error bars incorrectly. It appears as though they calculated their presentation times for each facilitation time correctly, however the error bars significantly change the interpretation of their results. Before, the authors incorrectly deduced that Presentation Time significantly differed as a function of Facilitation time. This claim is not true because of the large error bars.