cronbachs_alpha()

audpsychometric.cronbachs_alpha(ratings, *, axis=1)[source]

Calculate Cronbach’s alpha.

The Cronbach coefficient quantifying interrater agreement. Returns alpha as a float and additional information specific to this measure collated into a dictionary.

Cronbach’s alpha generalizes Cohen’s kappa and can handle three or more answers per variable. It is suitable for Likert type scale answers. A blogpost on congeneric reliability states that Cronbach’s alpha assumes essential tau-equivalence and underestimates reliability. A tau-equivalent measurement model is a special case of a congeneric measurement model with all loadings equal [cro].

A simplified formula is given in Hilsdorf [Hil] that relates the measure to the average reliability:

\[\alpha_{st} = \frac{N \times \bar{r}} {1 + (N - 1) \times \bar{r}}\]

where

  • \(N\) is the number of items (labelled chunks)

  • \(\bar{r}\) is the average correlation between the items

Parameters

  • ratings (Sequence) – ratings. When given as a 1-dimensional array, it is treated as a row vector

  • axis (int) – axis along which the rater confidence is computed. A value of 1 assumes stimuli as rows

Return type

Tuple[float, Dict]

Returns

Cronbach’s alpha and additional results lumped into dict