Interpolate gaps within trajectories

Use LOESS smoothing to fill in gaps of missing data within trajectories in a viewr object

Usage

fill_traj_gaps(
  obj_name,
  loess_degree = 1,
  loess_criterion = c("aicc", "gcv"),
  loess_family = c("gaussian", "symmetric"),
  loess_user_span = NULL
)

Arguments

obj_name

The input viewr object; a tibble or data.frame with attribute pathviewr_steps that includes "viewr". Trajectories must be predefined (i.e. via separate_trajectories()).

loess_degree

See "degree" argument of fANCOVA::loess.as()

loess_criterion

See "criterion" argument of fANCOVA::loess.as()

loess_family

See "family" argument of fANCOVA::loess.as()

loess_user_span

See "user.span" argument of fANCOVA::loess.as()

Value

A viewr object; a tibble or data.frame with attribute pathviewr_steps that includes "viewr" that now includes new observations (rows) as a result of interpolation to fill in missing data. A new column gaps_filled is added to the data to indicate original data ("No") vs data that have been inserted to fill gaps ("Yes").

Details

It is strongly recommended that the input viewr object be "cleaned" via select_x_percent() -> separate_trajectories() -> get_full_trajectories() prior to using this function. Doing so will ensure that only trajectories with minor gaps will be used in your analyses. This function will then enable you to interpolate missing data in those minor gaps.

Interpolation is handled by first fitting a series of LOESS regressions (via fANCOVA::loess.as()). In each regression, a position axis (e.g. position_length) is regressed against frame (frame is x-axis). From that relationship, values of missing position data are determined and then inserted into the original data set.

See loess.as for further details on parameters.

Author

Vikram B. Baliga

Examples

library(pathviewr)

## Import the example Motive data included in the package
motive_data <-
  read_motive_csv(system.file("extdata", "pathviewr_motive_example_data.csv",
                             package = 'pathviewr'))

## Clean, isolate, and label trajectories
motive_full <-
  motive_data %>%
  clean_viewr(desired_percent = 50,
              max_frame_gap = "autodetect",
              span = 0.95)
#> autodetect is an experimental feature -- please report issues.

## Interpolate missing data via this function
motive_filling <-
 motive_full %>%
 fill_traj_gaps()

## plot all trajectories (before)
plot_viewr_trajectories(motive_full, multi_plot = TRUE)

## plot all trajectories(after)
plot_viewr_trajectories(motive_filling, multi_plot = TRUE)