XTMEM2
Molly Lewis
2017-01-12
Replication of Spencer, Perone, Smith, and Samuelson (2011), Expt. 2 (sequential presentation of training items).
Read in raw data and anonymize
EXPTNUM <- 2
a.data.filename <- paste0("exp", EXPTNUM, "_A.csv")files = dir("../production-results/")
d = data.frame()
for (i in 1:length(files)[1]) {
s <- fromJSON(paste("../production-results/", files[i], sep = ""))
s$answers$asses = ifelse(is.null(s$answers$asses), "NA", s$answers$asses)
d = bind_rows(d, data.frame(s))
}
names(d) <- str_replace(names(d), "answers.", "")
d.anonymized <- anonymize.sids(d, "WorkerId")
write.csv(d.anonymized, a.data.filename)Munge
d2 = read.csv(a.data.filename)
d2.long = d2 %>%
gather(variable, value, contains("_")) %>%
mutate(trial_num = unlist(lapply(strsplit(as.character(variable),
"_T"),function(x) x[2])),
variable = unlist(lapply(strsplit(as.character(variable),
"_"),function(x) x[1]))) %>%
spread(variable, value) %>%
mutate(trial_num = as.numeric(trial_num)) %>%
mutate_if(is.character, funs(as.factor))
d2.munged = d2.long %>%
select(subids, trial_num, category, condition, selected) %>%
mutate(selected = lapply(str_split(selected, ","),
function(x) {str_sub(x, 4, 6)})) %>%
mutate(prop_sub = lapply(selected, function(x){sum(x == "sub")/2}),
prop_bas = lapply(selected, function(x){sum(x == "bas")/2}),
prop_sup = lapply(selected, function(x){sum(x == "sup")/4})) %>%
select(-selected)Reproduce XT2007a Figure 5.
ms2 = d2.munged %>%
gather(variable, value, c(prop_sub, prop_bas, prop_sup)) %>%
group_by(condition,variable) %>%
mutate(value = as.numeric(value)) %>%
multi_boot_standard(column = "value") %>%
mutate(variable = as.factor(variable))
ms.plot <- ms2
ms.plot$variable = factor(ms.plot$variable,levels(ms.plot$variable)[c(2,1,3)])
ms.plot$condition = factor(ms.plot$condition,levels(ms.plot$condition)[c(1,3,2,4)])
ms.plot$condition = plyr::mapvalues(ms.plot$condition,
from = c("one", "three_basic",
"three_subordinate",
"three_superordinate"),
to = c("1", "3 basic", "3 sub.", "3 super."))
ggplot(ms.plot, aes(x = condition, y = mean, group = variable, fill = variable)) +
geom_bar(position = "dodge", stat = "identity") +
geom_linerange(aes(ymin = ci_lower,
ymax = ci_upper),
position=position_dodge(width = .9)) +
ylab("Proportion of \ntest objects chosen") +
xlab("Examples") +
theme_bw() +
theme(legend.title = element_blank())
Here’s the XT2007a data:
grid.raster(readPNG("xtfig5.png"))
XTMEM 1 and 2 plotted as by SPSS:
d1 = read.csv("../../exp1/analysis/exp1_a.csv")
d1.long = d1 %>%
gather(variable, value, contains("_")) %>%
mutate(trial_num = unlist(lapply(strsplit(as.character(variable),
"_T"),function(x) x[2])),
variable = unlist(lapply(strsplit(as.character(variable),
"_"),function(x) x[1]))) %>%
spread(variable, value) %>%
mutate(trial_num = as.numeric(trial_num)) %>%
mutate_if(is.character, funs(as.factor))
d1.munged = d1.long %>%
select(subids, trial_num, category, condition, selected) %>%
mutate(selected = lapply(str_split(selected, ","),
function(x) {str_sub(x, 4, 6)})) %>%
mutate(prop_sub = lapply(selected, function(x){sum(x == "sub")/2}),
prop_bas = lapply(selected, function(x){sum(x == "bas")/2}),
prop_sup = lapply(selected, function(x){sum(x == "sup")/4})) %>%
select(-selected)
ms1 = d1.munged %>%
gather(variable, value, c(prop_sub, prop_bas, prop_sup)) %>%
group_by(condition,variable) %>%
mutate(value = as.numeric(value)) %>%
multi_boot_standard(column = "value") %>%
mutate(variable = as.factor(variable))
crit.conds1 <- ms1 %>%
filter(condition == "one" | condition == "three_subordinate") %>%
filter(variable == "prop_bas") %>%
mutate(exp = "XT2007a replication (E1)")
crit.conds2 <- ms2 %>%
filter(condition == "one" | condition == "three_subordinate") %>%
filter(variable == "prop_bas") %>%
mutate(exp = "SPSS replication (E2)")
crit.conds = rbind(crit.conds1, crit.conds2) %>%
mutate(exp = fct_rev(exp))
ggplot(crit.conds, aes(x = exp, y = mean, group = condition, fill = condition)) +
geom_bar(position = "dodge", stat = "identity") +
geom_linerange(aes(ymin = ci_lower,
ymax = ci_upper),
position=position_dodge(width = .9)) +
ylim(0,1)+
ylab("Proportion basic-level choices ") +
xlab("Experiment") +
theme_bw() +
theme(legend.title = element_blank())
And, the proportion basic for the 1 and sub conditions only in the SPSS data:
grid.raster(readPNG("spssfig3.png"))
We don’t see the flip observed by SPSS; looks almost identical to XTMEM1/XT2007a.
By category
ms2 = d2.munged %>%
gather(variable, value, c(prop_sub, prop_bas, prop_sup)) %>%
mutate(variable = as.factor(variable)) %>%
group_by(condition,variable,category) %>%
mutate(value = as.numeric(value)) %>%
multi_boot_standard(column = "value")
ms2$variable = factor(ms2$variable,levels(ms2$variable)[c(2,1,3)])
ms2$condition = factor(ms2$condition,levels(ms2$condition)[c(1,3,2,4)])
ms2$condition = plyr::mapvalues(ms2$condition,
from = c("one", "three_basic",
"three_subordinate", "three_superordinate"),
to = c("1", "3 basic", "3 sub.", "3 super."))
ggplot(ms2, aes(x = condition, y = mean, group = variable, fill = variable)) +
facet_grid(~category) +
geom_bar(position = "dodge", stat = "identity") +
geom_linerange(aes(ymin = ci_lower,
ymax = ci_upper),
position=position_dodge(width = .9)) +
ylab("Proportion of \ntest objects chosen") +
xlab("Examples") +
theme_bw() +
theme(legend.title = element_blank())
Post-task questions
d2 %>%
group_by(education) %>%
summarise(n = n()) %>%
kable()| education | n |
|---|---|
| 0 | 1 |
| 1 | 6 |
| 2 | 20 |
| 3 | 20 |
| 4 | 3 |
d2 %>%
group_by(enjoyment) %>%
summarise(n = n()) %>%
kable()| enjoyment | n |
|---|---|
| 1 | 17 |
| 2 | 33 |
d2 %>%
mutate(language = tolower(language)) %>%
group_by(language) %>%
summarise(n = n()) %>%
kable()| language | n |
|---|---|
| english | 49 |
| english | 1 |
d2 %>%
group_by(gender) %>%
summarise(n = n()) %>%
kable()| gender | n |
|---|---|
| 17 | |
| Female | 15 |
| Male | 18 |
d2 %>%
group_by(asses) %>%
summarise(n = n()) %>%
kable()| asses | n |
|---|---|
| Confused | 1 |
| Yes | 32 |
| NA | 17 |
d2 %>%
mutate(age = as.numeric(as.character(age))) %>%
ggplot(aes(x= age)) +
geom_histogram() +
theme_bw() +
ggtitle("Age distribution")
unique(d2$comments)## [1]
## [2] None
## [3] This was fun
## [4] interesting hit, thanks!
## [5] Great study
## [6] thanks
## [7] I like these, but it sucks never having feedback on them; like, was the turtle right to select? :P
## [8] That was fun!
## [9] N/A
## [10] I thought it had to do with they type, not the picture.
## [11] It was fun.
## [12] none
## [13] I enjoyed it! Thanks for the HIT! I did my best and hope the data is useful!
## [14] Good Luck!
## [15] A bit long for the pay, if I'm honest
## [16] I liked this hit - it was different and made me think about words generally vs specifically (i.e. a 'dog' vs a 'dalmatian'. Thanks and good luck
## [17] Thank you, good luck with your research!
## [18] Thanks!
## [19] ty
## [20] I enjoyed helping you with your research.
## [21] hi
## [22] It was interesting trying to decipher what was being asked. That is, in the early levels I had to choose (as I had very little information to go on) whether the word was specific or more generic in meaning.
## [23] This was so fun, I would love to do more if you have them
## 23 Levels: ... tyTask time
Just as a sanity check look at total task time: Expect participants to take longer in the SPSS experiment since its sequential. The data look consistent with ths.
d2 = mutate(d2, exp = "SPSS replication (E2)")
d1 = mutate(d1, exp = "XT2007a replication (E1)")
all = rbind(d2, d1)
all$SubmitTime = gsub("T|Z","",all$SubmitTime)
all$AcceptTime = gsub("T|Z","",all$AcceptTime)
all$SubmitTime = strptime(all$SubmitTime, "%F%T")
all$AcceptTime = strptime(all$AcceptTime, "%F%T")
all$total_time = as.numeric(all$SubmitTime) - as.numeric(all$AcceptTime)
all$exp = fct_rev(all$exp)
ggplot(all, aes(x = exp, y = total_time/60)) +
ylab("Task time (min)") +
geom_boxplot() +
theme_bw()
all %>%
select(-AcceptTime, -SubmitTime) %>%
group_by(exp) %>%
mutate(total_time = total_time/60) %>%
multi_boot_standard(column = "total_time") %>%
ggplot(aes(x = exp, y = mean, fill = exp)) +
geom_bar(position = "dodge", stat = "identity") +
geom_linerange(aes(ymin = ci_lower,
ymax = ci_upper),
position=position_dodge(width = .9)) +
ylab("Task time (min)") +
xlab("Examples") +
theme_bw() +
theme(legend.title = element_blank())