soc-xsit-fyp-analysis
Kyle MacDonald
August 16, 2014
Get the number of subjects in each condition.
d_df %>%
group_by(condition, intervalNum, numPicN) %>%
summarise(n_subs = n_distinct(subid))## Source: local data frame [32 x 4]
## Groups: condition, intervalNum
##
## condition intervalNum numPicN n_subs
## 1 No-Social 0 2 127
## 2 No-Social 0 4 114
## 3 No-Social 0 6 39
## 4 No-Social 0 8 117
## 5 No-Social 1 2 120
## 6 No-Social 1 4 118
## 7 No-Social 1 6 35
## 8 No-Social 1 8 114
## 9 No-Social 3 2 115
## 10 No-Social 3 4 117
## 11 No-Social 3 6 36
## 12 No-Social 3 8 114
## 13 No-Social 7 2 129
## 14 No-Social 7 4 115
## 15 No-Social 7 6 34
## 16 No-Social 7 8 114
## 17 Social 0 2 48
## 18 Social 0 4 82
## 19 Social 0 6 37
## 20 Social 0 8 43
## 21 Social 1 2 44
## 22 Social 1 4 88
## 23 Social 1 6 44
## 24 Social 1 8 44
## 25 Social 3 2 47
## 26 Social 3 4 87
## 27 Social 3 6 40
## 28 Social 3 8 43
## 29 Social 7 2 47
## 30 Social 7 4 90
## 31 Social 7 6 38
## 32 Social 7 8 38Exclude subjects who were choosing randomly during exposure trials
Flag trials on which subs chose the target of eye gaze.
#revalue face vid to match chosen indices
d_df$faceIdx6 <- revalue(d_df$face, c("eyes_left_90"=0, "eyes_right_90"=1,
"eyes_left"=2, "eyes_down_left"=3,
"eyes_down_right"=4, "eyes_right"=5,
"eyescenter"=-1))
d.expo_df <- d_df %>%
filter(exposureTrial == 1) %>%
mutate(correct_exposure = ifelse(numPic == 6,
chosenIdx == faceIdx6,
chosenIdx == faceIdx)) %>%
select(subid, itemNum, correct_exposure)Get test trials and merge with exposure trial information.
d.test_df <- d_df %>%
filter(testTrial == 1)
d.test_df <- join(d.expo_df, d.test_df, type = "full")## Joining by: subid, itemNumFlag subs in the social condition who performed worse than chance on exposure trials.
d.test_df <- d.test_df %>%
filter(condition == "Social") %>%
group_by(subid, numPic) %>%
summarise(mean_acc_exp = mean(correct_exposure)) %>%
mutate(include_expo = ifelse(numPic == 2 & mean_acc_exp > 0.5, 1,
ifelse(numPic == 4 & mean_acc_exp > 0.25, 1,
ifelse(numPic == 6 & mean_acc_exp > 0.17, 1,
ifelse(numPic == 8 & mean_acc_exp > 0.125, 1,
0))))) %>%
join(d.test_df, by = "subid", type = "full")Flag trials with extremely slow or fast RTs (+/- 2SD).
d.test_df <- d.test_df %>%
mutate(include_good_rt = ifelse(log(rt) > mean(log(rt)) + 2 * sd(log(rt)) |
log(rt) < mean(log(rt)) - 2 * sd(log(rt)),
0, 1))Experiment FYP: Social vs. No-social, Different delays, Different number of referents
Accuracy on exposure trials.
Accuracy on test trials, filtering out subjects who performed below chance on exposure trials.
ms_test_looks_filtered <- d.test_df %>%
filter(include_good_rt == 1,
(condition == "No-Social") |
(include_expo == 1 & correct_exposure == TRUE)
) %>%
group_by(condition, intervalNum, numPic, trialType) %>%
summarise(accuracy = mean(correct),
ci_low = ci.low(correct),
ci_high = ci.high(correct),
exclusionary_criteria = "subject-level")
acc_test_1 <- ggplot(data=ms_test_looks_filtered, aes(x=intervalNum, y=accuracy,
colour = condition)) +
geom_line(aes(linetype = trialType)) +
geom_errorbar(aes(ymin=accuracy - ci_low,
ymax=accuracy + ci_high), width = .1) +
geom_hline(aes(yintercept=1/numPic), linetype = "dashed") +
scale_x_continuous(limits=c(-0.5, 7.5), breaks=c(0,1,3,7)) +
scale_y_continuous(limits=c(0,1)) +
scale_colour_manual(values=c("firebrick1", "dodgerblue")) +
facet_grid(. ~ numPic) +
xlab("Intervening Trials") +
ylab("Proportion Correct") +
labs(colour = "Condition") +
labs(linetype = "Trial Type")Accuracy on test trials, without filtering subjects (only trial level filtering)
ms_test_looks_unfiltered <- d.test_df %>%
filter(include_good_rt == 1,
condition == "No-Social" | correct_exposure == TRUE) %>%
group_by(condition, intervalNum, numPic, trialType) %>%
summarise(accuracy = mean(correct),
ci_low = ci.low(correct),
ci_high = ci.high(correct),
exclusionary_criteria = "trial-level")
acc_test_2 <- ggplot(data=ms_test_looks_unfiltered, aes(x=intervalNum, y=accuracy,
colour = condition)) +
geom_line(aes(linetype = trialType)) +
geom_errorbar(aes(ymin=accuracy - ci_low,
ymax=accuracy + ci_high), width = .1) +
geom_hline(aes(yintercept=1/numPic), linetype = "dashed") +
scale_x_continuous(limits=c(-0.5, 7.5), breaks=c(0,1,3,7)) +
scale_y_continuous(limits=c(0,1)) +
scale_colour_manual(values=c("firebrick1", "dodgerblue")) +
facet_grid(. ~ numPic) +
xlab("Intervening Trials") +
ylab("Proportion Correct") +
labs(colour = "Condition") +
labs(linetype = "Trial Type")Now put both plots together.
