PG Reliability
Use for internal consistency and inter-rater/test-retest reliability.
Load
Read:
../../references/supervision-gates.md../../references/pingouin-api-quickref.md../../references/apa-output-template.md if writing results.
Function Choice
- Internal consistency of multiple scale items ->
pg.cronbach_alpha.
- Inter-rater, test-retest, or target-by-rater reliability ->
pg.intraclass_corr.
Required Inputs
- Item columns and reverse-coded item list for alpha.
- Whether items are all intended to measure one construct.
- Long-format ICC columns: target, rater, rating.
- ICC model decision: single vs average, consistency vs agreement, one-way vs two-way.
Code Patterns
Cronbach alpha:
items = df[["item1", "item2", "item3", "item4"]]
alpha, ci = pg.cronbach_alpha(data=items)
print({"alpha": round(alpha, 3), "ci95": ci})
Item-total screening:
for col in items.columns:
alpha_drop, ci_drop = pg.cronbach_alpha(data=items.drop(columns=col))
print(col, round(alpha_drop, 3), ci_drop)
ICC:
icc = pg.intraclass_corr(data=df, targets="target",
raters="rater", ratings="rating").round(3)
pg.print_table(icc)
Reporting
- Alpha: number of items, alpha, CI, and whether any reverse coding was applied.
- ICC: exact row/type from Pingouin output, ICC value, CI, F, df, p if available, and model interpretation.
Guardrails
- Alpha does not prove unidimensionality; mention factor structure if relevant.
- Low alpha can reflect few items, multidimensionality, poor items, or restricted range.
- Do not choose ICC row after seeing the most favorable value; pick model from design first.
- Make sure each target-rater pair is valid; duplicated ratings can corrupt ICC.
- End result-bearing answers with one compact S0-S5 audit line.
1---2name: pg-reliability3description: Run or generate Pingouin code for Cronbach alpha, item-scale reliability checks, and intraclass correlation for psychological ratings.4---56# PG Reliability78Use for internal consistency and inter-rater/test-retest reliability.910## Load1112Read:1314- `../../references/supervision-gates.md`15- `../../references/pingouin-api-quickref.md`16- `../../references/apa-output-template.md` if writing results.1718## Function Choice1920- Internal consistency of multiple scale items -> `pg.cronbach_alpha`.21- Inter-rater, test-retest, or target-by-rater reliability -> `pg.intraclass_corr`.2223## Required Inputs2425- Item columns and reverse-coded item list for alpha.26- Whether items are all intended to measure one construct.27- Long-format ICC columns: target, rater, rating.28- ICC model decision: single vs average, consistency vs agreement, one-way vs two-way.2930## Code Patterns3132Cronbach alpha:3334```python35items = df[["item1", "item2", "item3", "item4"]]36alpha, ci = pg.cronbach_alpha(data=items)37print({"alpha": round(alpha, 3), "ci95": ci})38```3940Item-total screening:4142```python43for col in items.columns:44 alpha_drop, ci_drop = pg.cronbach_alpha(data=items.drop(columns=col))45 print(col, round(alpha_drop, 3), ci_drop)46```4748ICC:4950```python51icc = pg.intraclass_corr(data=df, targets="target",52 raters="rater", ratings="rating").round(3)53pg.print_table(icc)54```5556## Reporting5758- Alpha: number of items, alpha, CI, and whether any reverse coding was applied.59- ICC: exact row/type from Pingouin output, ICC value, CI, F, df, p if available, and model interpretation.6061## Guardrails6263- Alpha does not prove unidimensionality; mention factor structure if relevant.64- Low alpha can reflect few items, multidimensionality, poor items, or restricted range.65- Do not choose ICC row after seeing the most favorable value; pick model from design first.66- Make sure each target-rater pair is valid; duplicated ratings can corrupt ICC.67- End result-bearing answers with one compact S0-S5 audit line.
Exekiel179