upxo.repqual.mcgs2d_representativeness_assesser module

Statistical representativeness assessment for 2D MCGS vs a target structure.

Class mc2repr runs distribution tests (e.g. KS, Mann–Whitney, Kruskal–Wallis, Pearson) on morphological properties of target vs sample grain structures. Targets may be EBSD-derived or UPXO MC/Voronoi GS. Results are stored for inspection; there is no single automatic accept/reject score in this module.

Import:

from upxo.repqual.mcgs2d_representativeness_assesser import mc2repr

class upxo.repqual.mcgs2d_representativeness_assesser.mc2repr(target_type=None, target=None, samples=None, par_bounds=None, metrics=None, kde_options=None, stest={'ks_p_threshold': 0.9, 'kw_p_threshold': 0.9, 'mw_p_threshold': 0.9, 'tests': ['correlation', 'kldiv', 'ks', 'jsdiv', 'mannwhitneyu', 'kruskalwallis']}, test_metrics=['mode0_location', 'mode0_count', 'mode1_location', 'mode1_count', 'mean'], parameters=['area'])

Bases: object

Statistical representativeness of 2D sample GS vs a target.

Runs distribution tests (KS, Mann–Whitney, Kruskal–Wallis, correlation, KL/JS divergence, etc.) on morphological properties of a target structure against one or more samples. Targets may be EBSD-derived or UPXO MC/Voronoi GS. Results are stored for inspection — there is no single automatic accept/reject score in this class.

target_type

Target source code:

• ebsd0 — unprocessed 2D EBSD (DefDAP)

• ebsd1 — processed DefDAP (e.g. remapped avg. orientation)

• umc2 / umc3 — UPXO Monte-Carlo 2D / 3D

• uvt2 — UPXO Voronoi tessellation 2D

• stats — morphology samples as dict or DataFrame columns

Type:

str

target

Target data: MCGS.gs[tslice], VTGS, DefDAP EBSD, or stats table.

samples

Sample name → grain-structure object, or 'make' to generate from an Excel dashboard (simulate → temporal slices → characterise).

Type:

dict

par_bounds

Per-property bounds, e.g. area/perimeter/aspect ratio → [[peak_loc_%], [peak_density_%], [JS_bounds]].

Type:

dict

metrics

Qualification metrics (e.g. modes_n, modes_loc, skewness).

Type:

list

kde_options

KDE options (bw_method: 'scott', 'silverman', or scalar).

Type:

dict

stest, test_metrics, performance

Configured statistical tests and computed outcomes.

target_type

target

samples

par_bounds

metrics

kde_options

stest

test_metrics

parameters

performance

load_target(target=None, target_type=None)

Load or import target.

load_samples(samples=None)

Load or import samples.

add_sample(sample=None)

Add or insert sample.

set_stests(tests)

Set or update stests.

set_cor_thresh(cor_threshold)

Set or update cor thresh.

set_kldiv_thresh(kldiv_thresh)

Set or update kldiv thresh.

set_ks_thresh(ks_thresh_D, ks_thresh_P)

Set or update ks thresh.

set_jsdiv_thresh(jsdiv_thresh)

Set or update jsdiv thresh.

prop_to_excel(filename='pxtal_properties')

Prop to excel.

build_distribution_dataset()

Build and return distribution dataset.

determine_distr_type()

Determine distr type.

test()

TEST 1: correlation: For two datasets, it is a measure of the linear relationship between them. If correlation is close to 1 then, the distributions are very similar.

TEST 2: kldiv:

TEST 3: ks: Kolmogorov-Smirnov test: Determines of the two distribution samples differ significantly. It uses cumulative distributions of the two datasets. Retyurns D-statistic and P-value.

• D-statistic: maximum absolute difference of the cumulative

distributions (absolute max distance (supremum) b/w the CDFs of the two samples). A smaller D-static value is indicative of similar distributions. * P-value: probability that thwe tywo distributions are similar. If p-value is low (<= 0.05), distributions are different. If p-value is high (> 0.05), we cannot reject the null-hypothesis that the two distributions are the same. * Note: if P <= 0.05: the null hypothesis that the two samples are drawn from tyhe sample sample can be rejected, indicating that the samples are not representative of the target

TEST 4: jsdiv: P value will allways be between 0 and 1. @ 0: Distributions are identical. @ 1: Distributions are completely different

TEST 5: mannwhitneyu: Mann-Whitney test: Used to determine if two ‘ distribution samples are drawn from a population having the same population. If P-value is less than or equal to 0.05, then different distributiopns. If P-value is > 0.05, then the two disrtirbutions are similar.

TEST 6: kruskalwallis: Kruskal-wallis test. Used to determine if there are statistically significant differences between two distributions.

stat_tests

test_threshold

distr_type