Talk:Krippendorff's alpha

The notation in here is extremely awkward and makes the discussion difficult to follow. Would anyone object to a massive re-write? Ken K (talk) 16:19, 3 December 2014 (UTC)[reply]

Many changes implemented. I believe this is a substantial improvement, as in the previous version, it seems that the same symbols had several different meanings. I also included mathematical expressions where only English had been used previously.

HOWEVER. I lack the knowledge to complete the "Embrace of other statistics section" accurately. This section does need editing; it it still quite unclear. Ken K (talk) 23:33, 4 December 2014 (UTC)[reply]

There are still huge numbers of undefined or ill-defined terms here in this article. By way of example neither $n_{c}$ nor $n_{k}$ is defined before it is used. There are many other such cases.

"When four or more coders are involved, o_ck may be fractions"

The following example disproves this.

Units u:	1	2	3	4	5
Coder A	3	2	1	1	2
Coder B	3	1	2	2	2
Coder C	3	2	1	3	2

The coincidence matrix for this counterexample is:

Values v or v' :	1	2	3	n_v
Value 1	1	$2{\frac {1}{2}}$	${\frac {1}{2}}$	4
Value 2	$2{\frac {1}{2}}$	4	${\frac {1}{2}}$	7
Value 3	${\frac {1}{2}}$	${\frac {1}{2}}$	3	4
Frequency n_v'	4	7	4	15

With two raters no fractions can occur, as the denominator will always be 1. Suggest that the sentence is changed to: "...when three or more coders are involved..." --EelcoVriez (talk) 09:27, 13 November 2015 (UTC)[reply]

Comment on the computational example

There is no hint what $\delta _{metric}$ is. It is not mentioned above and also not in the linked article. I lack the knowledge to complete this. — Preceding unsigned comment added by 46.193.65.129 (talk) 08:58, 30 March 2017 (UTC)[reply]

Error in Computational Example?

It's noted that $\delta (v,v')=\delta (v',v)$ so the author only includes one of the cases in the numerator, but shouldn't each one be multiplied by 2? If that is not correct, why not? StatsJunkie (talk) — Preceding undated comment added 23:05, 28 December 2020 (UTC)[reply]

References incomplete

When trying to find the primary source for the 0.667/0.8 thresholds I noticed the 2004 publication by Krippendorf is not in the references. It seems to refer to "Content analysis: An introduction to its methodology (2nd ed.)". Muxarin (talk) 14:33, 17 November 2020 (UTC)[reply]