gistfile1.txt

```{r}
library(tictoc)
library(splithalf)
library(splithalfr)
```


```{r}
tic()
data("ds_vpt", package = "splithalfr")

# ds_vpt <- subset(ds_vpt, block_type == "assess")  # cancelling this line as ds_vpt contains assess trials only

vpt_fn_sets <- function (ds) {
  return (list(
    # Probe-at-test
    patt_yes = subset(ds, patt == "yes"),
    # Probe-at-control
    patt_no  = subset(ds, patt == "no")
  ))
}

vpt_fn_score <- function (sets) {
  rt_yes <- subset(sets$patt_yes, response == 1)$rt
  rt_yes <- rt_yes[rt_yes >= 200 & rt_yes <= 520]
  rt_no <- subset(sets$patt_no, response == 1)$rt
  rt_no <- rt_no[rt_no >= 200 & rt_no <= 520]
  return (mean(rt_no) - mean(rt_yes))
}

## splithalfing

vpt_splits <- sh_apply(ds_vpt, "UserID", vpt_fn_sets, vpt_fn_score, 5000)


# Spearman-Brown
mean_sb_by_split(vpt_splits)


toc()
```

```{r}
tic()

ds_vpt <- as.data.frame(splithalfr::ds_vpt)

ds_vpt$correct [ds_vpt$response == 1] <- 1
ds_vpt$correct [ds_vpt$response != 1] <- 0

ds_vpt$rt <- as.numeric(ds_vpt$rt)
ds_vpt$rt <- as.numeric(ds_vpt$rt)

splithalf_diff(ds_vpt, RTmintrim = 200, RTmaxtrim = 520, no.iterations = 5000, incErrors = F, var.RT = "rt", var.participant = "UserID", var.correct = "correct",  var.trialnum = "trial",  var.compare = "patt", compare1 = "yes", compare2 = "no") 

toc()

```


```{r}

# check if difference in estimates may be due to splithalf() using an adapted Spearman-Brown formula in which the absolute value of the correlation is used - this makes it so that a negative correlation between two parts also results in an SB estimate that correspons to the SB estimate if the correlation had been positive:
# example:
corpos <- .2
corneg <- -.2

2*corpos/(1+corpos) # 0.3333333
2*corneg/(1+corneg) # -0.5
2*corneg/(1+ abs(corneg)) #  -0.3333333

# adapt formula's from splithalfr() to implement the abs(cor(x,y)) in the SB function:

spearman_brownabs <- function (x, y) {
  return (2 * cor(x,y) / (1 + abs(cor(x,y))))
}

mean_sbabs_by_split <- function (ds) {
  return (mean_rel_by_split(ds, spearman_brownabs))
}

require(dplyr)
mean_rel_by_split <- function (ds, fn_rel) {
  # Check on missing values
  if (any(is.na(ds[c("score_1", "score_2")]))) {
    warning ("input data contained missing values; these were pairwise removed before calculating the reliability coefficient")
    # Remove missing values pairwise
    rows_missing = is.na(ds$score_1) | is.na(ds$score_2)
    ds = ds[!rows_missing,]
  }
  ds_rs <- ds %>%
    group_by(split) %>%
    summarize(
      r = fn_rel(.data$score_1, .data$score_2)
    )
  return (mean(ds_rs$r))
}

# compare SB estimates with and without abs(cor(x,y)) tweak: 

# Spearman-Brown
mean_sb_by_split(vpt_splits)
# Spearman-Brown
mean_sbabs_by_split(vpt_splits)

# the latter one is more in line with the SB esitmate returned by splithalf()
```