Dear all,
I've run into an issue that I can't for the life of me figure out. I have a fairly complicated Inquisit script that runs an approach-avoidance manikin task in trial pairs, where the condition assignment depends on 8 two-level factors (i.e., 256 conditions). In addition, the script needs to do a lot of other stuff (e.g., randomly select color sets for targets, measure slow responses and incorrect responses and assign a reward based on quotas, include a practice block with a separate condition list, repeat incorrect trials until the participant gets them right in the practice block, but only once in the main task etc.), which has made it difficult to debug this problem.
The observed behavior is that when I run my script file using the test monkey, the generated test data will occasionally have irregular distributions of conditions. As an example, I've uploaded a data file I generated using a simplified version of the experiment with 128 conditions (also attached as a script). First, one should note that the data file contains data from the practice block as well (all blocks where BlockCounter < 1). It's also important to note that there are more than the expected 256 trials (128 trial pairs, 1 per condition = 256 trials) in the data file, as this script also includes code that causes error trials to repeat once at the end of the block. Thus, the expected behavior of the script is to run 128 trial pairs in a block, drawing from the Conditions <list> in the script (which has 128 elements), and then to continue the block with single repetitions of conditions that were responded to incorrectly (captured in the ConditionsCollectedErrors <list> at the end of the AAT trial). This should mean that in the first 256 trials, the conditions (indexed in values.SequenceType) should each occur precisely once.
As can be seen in the data file, this is not the case. Subject number 5 sees one sequence twice in blocks 2, 3, and 4 (not always the same sequence). The SequenceType for the last trial pair among the first 128 (i.e. trials 255+256) in these blocks occur twice in that set of 128 trial pairs instead of once, with another SequenceType presumably conversely missing in the set. It's almost as though list.Conditions is selected from one too many times and the first selection is being dropped somehow. Unfortunately, as can also be seen in the dataset, this does not happen predictably - of 15 monkeys in this dataset, only 1 shows the issue.
I've tried reducing the problem to a simpler script, but the bug doesn't appear to occur if I limit list.Conditions to the first 8, 16, 32, or 64 elements (admittedly I only tested each of these with 15 monkeys, so I might just have failed to detect it). I'm relatively sure that the problem lies somewhere with my additional bells and whistles (intuitively, I particularly suspect the code that repeats the error trials), but I've gone over it multiple times and can't find the issue.
I would appreciate any help or tips as to what might be going on!
I couldn't upload the data file to the forum because of its size (8mb), so here it is on Sync:
https://ln5.sync.com/dl/4034d7ec0/8w24j3cf-nmxnc22y-viay2gpq-v92pchgeAlso, here's the R code I use to check whether the conditions are being assigned properly (requires tidyverse):
library(tidyverse)
rawdatasingles <- read_tsv("X:/ALLG2/AAT SR-Binding/Experiment 6/Programming/data/Simplified 128 conditions_raw.iqdat")
# getting prime and probes into one row
rawdata <- rawdatasingles %>% mutate(TrialCounter = if_else(TrialCounter %% 2 == 0,TrialCounter-1,TrialCounter)) %>%
mutate(BlockCounter = case_when(BlockCounter == 0 ~ PracticeBlocksThatWereCompleted * -1,
TRUE ~ BlockCounter)) %>%
pivot_wider(id_cols = c(subject,BlockCounter, TrialCounter,DistracterCondition,GoalResponseCondition,
MotorResponseCondition,RelevantStimulusCondition), names_from = CurrentTrialTypeInSequence,
values_from = c(DesiredResponse,ManikinPosition,TargetStimColor,
AATTargetWord,correct,response,SequenceType))
# checking whether conditions are evenly assigned
rawdata %>%
filter(BlockCounter > 0 & TrialCounter <= 256) %>%
count(subject,BlockCounter,DistracterCondition,GoalResponseCondition,MotorResponseCondition,RelevantStimulusCondition,
DesiredResponse_Prime,ManikinPosition_Prime,TargetStimColor_Prime,AATTargetWord_Prime) %>%
arrange(desc(n)) %>%
group_by(subject,BlockCounter,n) %>%
count(GoalResponseCondition) %>% # vary the condition variable in this line to see distributions across other condition combinations
view
rawdata %>%
filter(BlockCounter > 0 & TrialCounter <= 256) %>%
count(subject,BlockCounter,SequenceType_Prime) %>%
arrange(desc(n)) %>%
group_by(subject,BlockCounter,n) %>%
view