Skip to contents

AOI-based gaze interpolation

interpolate_looks.Rd

Fills in windows of missing data if the same AOI is fixated at beginning and end of the missing data window. For example, the sequence "Target", NA, NA, "Target" would be interpolated to be "Target", "Target", "Target", "Target".

Usage

interpolate_looks(
  x,
  window,
  fps,
  response_col,
  interp_col,
  fillable,
  missing_looks
)

Arguments

x

a dataframe of grouped eyetracking data. Each row should be a single frame of eyetracking. Use dplyr::group_by() to set the grouping columns for the data. The groups should specify a single trial of eyetracking data.

window

maximum amount of missing data (milliseconds) that can be interpolated. Only spans of missing data with less than or equal to this duration will be interpolated

fps

number of eyetracking frames (dataframe rows) per second

response_col

(character) name of the column with the eyetracking response data

interp_col

(character) name of a column to add to the dataframe. This column records whether each frame was interpolated (TRUE) or not (FALSE)

fillable

values in the response column where interpolation is legal. These would typically be AOI locations.

missing_looks

values that can be imputed.

Details

Use window to constrain the duration of missing data windows that can be filled. We conventionally use 150ms because we would not expect someone to shift their gaze from Image A to Image B to Image A in that amount of time.

Examples

# We have time in ms, measured at 60 fps, and
# we want to fill in gaps of 100 ms.
looks <- tibble::tribble(
  ~Subject, ~Trial, ~Time,    ~AOI,         ~Hint,
       "A",      1,  1000,  "Left",     "present",
       "A",      1,  1017,  "Left",     "present",
       "A",      1,  1034,      NA,   "legal gap",
       "A",      1,  1051,      NA,   "legal gap",
       "A",      1,  1068,      NA,   "legal gap",
       "A",      1,  1084,  "Left",     "present",
       "A",      1,  1100,      NA, "illegal gap",
       "A",      2,   983,  "Left",     "present",
       "A",      2,  1000, "Right",     "present",
       "A",      2,  1017,      NA, "illegal gap",
       "A",      2,  1034,      NA, "illegal gap",
       "A",      2,  1051,      NA, "illegal gap",
       "A",      2,  1068,      NA, "illegal gap",
       "A",      2,  1084,      NA, "illegal gap",
       "A",      2,  1100,      NA, "illegal gap",
       "A",      2,  1118,      NA, "illegal gap",
       "A",      2,  1135, "Right",     "present",
)

# Note that only the "legal gap" rows were interpolated
looks |>
  dplyr::group_by(Trial) |>
  interpolate_looks(
    window = 100,
    fps = 60,
    response_col = "AOI",
    interp_col = "Interpolated",
    fillable = c("Left", "Right"),
    missing_looks = NA
  )
#> # A tibble: 17 × 6
#> # Groups:   Trial [2]
#>    Subject Trial  Time AOI   Hint        Interpolated
#>    <chr>   <dbl> <dbl> <chr> <chr>       <lgl>       
#>  1 A           1  1000 Left  present     FALSE       
#>  2 A           1  1017 Left  present     FALSE       
#>  3 A           1  1034 Left  legal gap   TRUE        
#>  4 A           1  1051 Left  legal gap   TRUE        
#>  5 A           1  1068 Left  legal gap   TRUE        
#>  6 A           1  1084 Left  present     FALSE       
#>  7 A           1  1100 NA    illegal gap FALSE       
#>  8 A           2   983 Left  present     FALSE       
#>  9 A           2  1000 Right present     FALSE       
#> 10 A           2  1017 NA    illegal gap FALSE       
#> 11 A           2  1034 NA    illegal gap FALSE       
#> 12 A           2  1051 NA    illegal gap FALSE       
#> 13 A           2  1068 NA    illegal gap FALSE       
#> 14 A           2  1084 NA    illegal gap FALSE       
#> 15 A           2  1100 NA    illegal gap FALSE       
#> 16 A           2  1118 NA    illegal gap FALSE       
#> 17 A           2  1135 Right present     FALSE