Detect periodicity in a time series — detect_periodicity • hmisindia

Uses spectral analysis (periodogram) and autocorrelation to determine if a numeric time series has periodic/seasonal patterns.

Usage

detect_periodicity(
  x,
  value_col = "value",
  time_col = "monyear",
  detrend = TRUE
)

# S3 method for class 'hmis_periodicity'
plot(x, ...)

Arguments

x: An hmis_periodicity object from detect_periodicity().
value_col: Column name for values if x is a data frame. Default "value".
time_col: Column name for time if x is a data frame. Default "monyear". Parsed via parse_monyear() if character.
detrend: Logical. Remove linear trend before analysis. Default TRUE.
...: Additional arguments (ignored).

Value

A list with class "hmis_periodicity" containing:

is_periodic: Logical. TRUE if significant periodicity detected.
dominant_period: Dominant period in number of observations (e.g., 12 for monthly data with annual cycle).
spectral_power: Relative spectral power at the dominant frequency.
acf_peak: Autocorrelation at the dominant lag.
frequencies: Data frame of top frequencies with periods and power.
spectrum: The raw spectrum object from stats::spec.pgram().

Examples

if (FALSE) { # \dontrun{
births <- get_hmis("Number of female live births",
  state = "Kerala", category = "Total", sector = "Total"
)
result <- detect_periodicity(births)
print(result)
plot(result)
} # }