Skip to content

Statistics & agreement

Beyond building a chronology, dplPy provides the descriptive statistics, signal- strength measures, and agreement tests used to characterize a collection and judge whether it carries a common, datable signal.

Descriptive statistics: stats

stats reports per-series summary statistics — span, mean, median, standard deviation, skew, kurtosis, the Gini coefficient, and first-order autocorrelation:

import dplpy as dpl

rwl = dpl.readers("ca533.rwl", header=True)
dpl.stats(rwl).head(3)

   series  first  last  year   mean  median  stdev   skew  kurtosis   gini    ar1
1  CAM011   1530  1983   454  0.440    0.40  0.222  1.029     1.102  0.273  0.696
2  CAM021   1433  1983   551  0.424    0.40  0.185  0.946     1.110  0.237  0.701
3  CAM031   1356  1983   628  0.349    0.29  0.214  0.690    -0.366  0.341  0.808

ar1 here is the autocorrelation-function coefficient at lag 1 (matching dplR's rwl.stats), not an OLS AR(1) slope.

Mean sensitivity: sens1 and sens2

Mean sensitivity measures year-to-year variability. sens1 is the classic Douglass measure; sens2 is the trend-robust variant (Biondi & Qeadan 2008). Both take a single series:

s = rwl["CAM011"].dropna()
dpl.sens1(s)   # 0.3437
dpl.sens2(s)   # 0.2960

As in dplR, sensitivity is deliberately kept out of the stats table and exposed as its own functions.

Signal strength: rwi_stats, sss

rwi_stats computes the population-signal statistics on detrended series — inter-series correlation (rbar), the expressed population signal (EPS), and the signal-to-noise ratio (SNR). It groups cores by tree using an ids table from read_ids:

rwi = dpl.detrend(rwl, fit="Spline")
ids = dpl.read_ids(rwl)
dpl.rwi_stats(rwi, ids)

   n_cores  n_trees   n  n_tot  ...  rbar_eff    eps     snr
0       34       34  34    523  ...     0.423  0.961  24.875

rwi_stats_running computes the same statistics in a moving window to see how signal strength changes through time. sss (subsample signal strength) estimates how well a reduced sample reproduces the full-collection signal — useful for deciding how far back a chronology remains reliable.

Agreement: glk and sgc

glk (Gleichläufigkeit) and sgc (synchronous growth changes) measure the proportion of intervals in which two series move in the same direction — a classic non-parametric agreement test. Both return a dictionary of pairwise values and their significance:

dpl.glk(rwl, overlap=50)
dpl.sgc(rwl, overlap=50)

The p-values follow dplR's formulation (a p-value can exceed 1 when the statistic is below 0.5).

Tree-level and interval helpers

  • tree_mean averages cores to the tree level first (using an ids table), so multi-core trees are not over-weighted in downstream statistics.
  • common_interval finds the interval or subset of series that maximizes overlap (by years, by series, or both) — a faithful port of dplR's common.interval, handy before computing statistics that need complete data.

Basal area increment: bai_out and bai_in

bai_out and bai_in convert ring widths to basal area increment, working inward from the outermost ring (bai_out) or outward from the pith (bai_in, which can take a pith-to-first-ring distance):

dpl.bai_out(rwl)
dpl.bai_in(rwl, d2pith=pith_distances)

See the Statistics & indices reference for the full parameter lists.