Diagnostic Summary for Time Series Objects

Provides a compact diagnostic summary for univariate ts objects, extending classical descriptive statistics with key time series diagnostics.

Usage

# S3 method for class 'ts'
desc(
  x,
  maxLag = 12,
  main = NULL,
  plotit = NULL,
  verbose = NULL,
  digits = NULL,
  ...
)

# S3 method for class 'Desc.ts'
print(x, digits = NULL, ...)

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

Arguments

x: a univariate object of class "ts"
maxLag: number of lags used in the Ljung-Box test; defaults to 12
main: character string, NULL, or NA, defining the main title. By default (main = NULL) the title will be composed as: (<class(es)>). If NA, no title is printed.
plotit: logical. Should a plot be created? The plot type depends on the classes of the variables. Default can be defined by the option plotit, if it does not exist then it's set to FALSE.
verbose: integer controlling verbosity of table output. One of 1 (minimal), 2 (default), 3 (extensive). Applies to tables only.
digits: number of digits used to format numeric values
...: further arguments passed to methods

Value

an object of class c("Desc.ts", "Desc") with, among others, the components acf1, ljungbox, adf, kpss (test results), stationary (logical, the combined rule above), trend (named vector slope, p.value) and boxcoxlambda

Details

The function reports:

Lag-1 autocorrelation
Ljung-Box test for overall autocorrelation
Augmented Dickey-Fuller (ADF) test
KPSS test
Linear trend estimation (slope and p-value)
Suggested Box-Cox transformation parameter

The goal is to provide quick diagnostic guidance before model fitting (e.g., ARIMA specification).

Stationarity is evaluated using both the Augmented Dickey-Fuller (ADF) and KPSS tests. A combined decision rule is used: the series is considered stationary if the ADF test rejects the null hypothesis of a unit root (p < 0.05) and the KPSS test does not reject the null hypothesis of stationarity (p > 0.05).

The Box-Cox transformation parameter is estimated using boxCoxLambda().

Missing values are allowed. The autocorrelation and the Ljung-Box test keep the time structure (na.action = na.pass), the ADF and KPSS tests and the Box-Cox parameter use the observed values only. The Box-Cox parameter is NA unless all values are strictly positive.

References

Box, G. E. P., Jenkins, G. M., Reinsel, G. C., & Ljung, G. M. (2015). Time Series Analysis: Forecasting and Control.

Hyndman, R. J., & Athanasopoulos, G. (2021). Forecasting: Principles and Practice.

Examples

desc(AirPassengers)
#> ────────────────────────────────────────────────────────────────────────────── 
#> AirPassengers (ts)
#> 
#> length        n        NAs  unique    0s
#>    144      144          0     118     0
#>          100.0%       0.0%          0.0%
#>                                         
#>  start      end  frequency              
#> 1949-1  1960-12         12              
#> 
#> lag-1 autocorrelation : 0.948
#> Ljung-Box (lag 12)     : Q = 1036.482, p = < 0.001
#> ADF                   : -0.352, p = 0.502
#> KPSS                  : 2.739, p = <0.010
#> stationary            : no
#> linear trend          : slope = 31.886, p = < 0.001
#> Box-Cox lambda        : 0.111
#> 

desc(Nile, maxLag = 10)
#> ────────────────────────────────────────────────────────────────────────────── 
#> Nile (ts)
#> 
#> length       n        NAs  unique    0s
#>    100     100          0      85     0
#>         100.0%       0.0%          0.0%
#>                                        
#>  start     end  frequency              
#> 1871-1  1970-1          1              
#> 
#> lag-1 autocorrelation : 0.498
#> Ljung-Box (lag 10)     : Q = 88.127, p = < 0.001
#> ADF                   : -0.964, p = 0.306
#> KPSS                  : 0.965, p = <0.010
#> stationary            : no
#> linear trend          : slope = -2.714, p = < 0.001
#> Box-Cox lambda        : 0.999
#>