Skip to contents

Apply the rolling standard deviation test to current variables

adcp_test_rolling_sd.Rd

The rolling standard deviation is calculated using the following algorithm:

Usage

adcp_test_rolling_sd(
  dat,
  current_rolling_sd_table = NULL,
  county = NULL,
  vars = NULL,
  period_hours = 12,
  max_interval_hours = 2,
  align_window = "center",
  keep_sd_cols = FALSE
)

Arguments

dat

Data frame of current data in wide format.

current_rolling_sd_table

Data frame with at least two columns: variable: must match the names of the variables being tested in dat. rolling_sd_max: maximum accepted value for the rolling standard deviation.

Default values are used if rolling_sd_table = NULL. To see the default rolling_sd_table, type subset(current_thresholds, qc_test == "rolling_sd") in the console.

county

Character string indicating the county from which dat was collected. Used to filter the default rolling_sd_table. Not required if there is a county column in dat.

vars

Vector of character strings indicating which columns to pivot. Default is all variables.

period_hours

Length of a full cycle in hours. Default assumes a tidal cycle of period_hours = 12.

max_interval_hours

Maximum accepted interval between two observations. If the interval between two observations is greater than this value, the rolling standard deviation will be set to NA. This is important because for large intervals, the number of observations in period_hours will be small. For example, if samples are collected every 6 hours, only 2 observations would be used to calculate roll_sd.

align_window

Alignment for the window used to calculate the rolling standard deviation. Passed to zoo::rollapply(). Default is align_window = "center". Other options are "right" (backward window) and "left" (forward window).

keep_sd_cols

Logical value. If TRUE, the columns used to produce the rolling standard deviation (int_sample, n_sample, and sd_roll) are returned in dat. Default is FALSE.

Value

Returns dat in a wide format, with rolling standard deviation flag columns for each variable in the form "rolling_sd_flag_variable".

Details

1. Calculate the interval between the current and previous observation (int_sample; in minutes).

2. Determine the number of samples in the specified time frame (n_sample = round((60 / int_sample) * period_hours))

3. If int_sample is really large (> max_interval_hours), then set the effective sample number (n_sample_effective) to 0.

4. Calculate the rolling standard deviation, with the width = n_sample_effective and alignment specified by align_window. sd_roll is rounded to 2 decimal places.

When int_sample = period_hours, n_sample = 1, so sd_roll = NA.

See also

Other tests: adcp_test_spike()