Calculate fire intervals from a composite
intervals(comp, densfun = "weibull")A composite instance, usually output from composite().
Should contain only one series.
String giving desired distribution to fit. Either "weibull" or "lognormal". Default is "weibull".
An intervals object. intervals have components:
"intervals" an integer vector giving the actual fire intervals.
"fitdistr" a fitdistr object from MASS::fitdistr() representing the
density function fit.
"densfun" a string giving the name of the density function used.
"kstest" an htest object from stats::ks.test() giving the result of a
one-sample Kolmogorov-Smirnov test.
"shapirotest" an htest object from stats::shapiro.test() giving the
result of a Shapiro-Wilk normality test.
"comp_name" a string giving the name of the interval's input composite.
"event_range" an integer vector giving the year range (min, max) of events used to create this intervals.
composite() to create a composite object.
mean.intervals() gets mean fire interval.
median.intervals() gets median fire interval.
quantile.intervals() get fit distribution quantiles.
plot_intervals_dist() plots intervals.
min.intervals() gives the minimum fire interval.
max.intervals() gives the maximum fire interval.
print.intervals() prints common fire-interval summary statistics.
data(pgm)
interv <- intervals(composite(pgm))
#> Warning: ties should not be present for the Kolmogorov-Smirnov test
print(interv)
#> Interval Analysis
#> =================
#>
#> Composite name: COMP
#> Events range: 1636 to 1871
#>
#> Total intervals: 20
#> Mean interval: 11.8
#> Median interval: 10
#> Weibull median: 10.6
#> Standard deviation: 7.7
#> Minimum interval: 2
#> Maximum interval: 30
#>
#>
#> Theoretical distribution
#> ------------------------
#>
#> Fit distribution: weibull
#>
#> shape scale
#> 1.6565206 13.2125358
#> ( 0.2797517) ( 1.8871133)
#>
#>
#> Percentiles
#> -----------
#>
#> lower (12.5%): 3.9 | weibull median (50.0%): 10.6 | upper (87.5%): 20.6
#>
#> Asymptotic one-sample Kolmogorov-Smirnov test
#> ---------------------------------------------
#>
#> D^- = 0.12776, p = 0.89978
#> Null hypothesis: The intervals were sampled from the theoretical
#> distribution.
#> Alt. hypothesis: The intervals were not sampled from the theoretical
#> distribution.
#>
#>
mean(interv) # Mean interval
#> [1] 11.75
# Now fit log-normal distribution instead of Weibull.
intervals(composite(pgm), densfun = "lognormal")
#> Warning: ties should not be present for the Kolmogorov-Smirnov test
#> Interval Analysis
#> =================
#>
#> Composite name: COMP
#> Events range: 1636 to 1871
#>
#> Total intervals: 20
#> Mean interval: 11.8
#> Median interval: 10
#> Weibull median: 10.6
#> Standard deviation: 7.7
#> Minimum interval: 2
#> Maximum interval: 30
#>
#>
#> Theoretical distribution
#> ------------------------
#>
#> Fit distribution: lognormal
#>
#> meanlog sdlog
#> 2.2538697 0.6779964
#> (0.1516046) (0.1072006)
#>
#>
#> Percentiles
#> -----------
#>
#> lower (12.5%): 4.4 | lognormal median (50.0%): 9.5 | upper (87.5%): 20.8
#>
#> Asymptotic one-sample Kolmogorov-Smirnov test
#> ---------------------------------------------
#>
#> D^- = 0.09849, p = 0.99016
#> Null hypothesis: The intervals were sampled from the theoretical
#> distribution.
#> Alt. hypothesis: The intervals were not sampled from the theoretical
#> distribution.
#>
#>
if (FALSE) {
# Boxplot of fire interval distribution.
boxplot(intervals(composite(pgm))$intervals)
}