Comparison of real data and simulated data analysis based on the standard error of measurement for stopping algorithm in a computerized adaptive testing: a psychometric study

[ben-domingue]

just the real data

https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/R5STH3

[saviranadela]

data:
CAT_DongGiSeo_2024.csv

code:

library(readxl)
library(tidyr)
library(dplyr)

df2 <- read_xlsx('jeehp-21-18-suppl2.xlsx')
df3 <- read_xlsx('jeehp-21-18-suppl3.xlsx')

df3 <- df3 %>%
  mutate(joined_indicator = TRUE)

merged <- df2 %>%
  left_join(df3, by = c("taker_id", "item_code", "is_correct")) %>%
  mutate(SEM_type = ifelse(is.na(joined_indicator), "0.25", "0.25_0.3")) %>%
  select(taker_id, item_code, SEM_type, is_correct)

merged <- merged %>%
  rename(id = taker_id, item = item_code, resp = is_correct) %>%
  mutate(resp = recode(resp, "Y" = 1, "N" = 0))

write.csv(merged, "CAT_DongGiSeo_2024.csv", row.names=FALSE)

cc: @KingArthur0205

[saviranadela]

i combined the items from both SEM 0.3 and SEM 0.25 datasets and added a SEM_type column to distinguish which are the additional items from SEM 0.25. additionally, i captured the correct/incorrect as the responses. please let me know if you have any feedback!

[saviranadela]

PR #689

[ben-domingue]

hm. this one is kind of murky. given that they are using CAT there are very few responses/item. i'm tentatively inclined to leave out for the moment as it isn't that much data and will require flags to ensure that it gets managed appropriately. thoughts @saviranadela @KingArthur0205 ? sorry about this one; hard to know it was going to be a problem until i saw the very sparse response matrix!

[ben-domingue]

thoughts about this? might be that we should just close and leave out of irw. i'm going to close now but we can reopen if anyone disagrees.