Cohen's kappa
Encyclopedia
Cohen's kappa coefficient is a statistical measure of inter-rater agreement or inter-annotator agreement for qualitative (categorical) items. It is generally thought to be a more robust measure than simple percent agreement calculation since κ takes into account the agreement occurring by chance. Some researchers (e.g. Strijbos, Martens, Prins, & Jochems, 2006) have expressed concern over κ's tendency to take the observed categories' frequencies as givens, which can have the effect of underestimating agreement for a category that is also commonly used; for this reason, κ is considered an overly conservative measure of agreement.
Others (e.g., Uebersax, 1987) contest the assertion that kappa "takes into account" chance agreement. To do this effectively would require an explicit model of how chance affects rater decisions. The so-called chance adjustment of kappa statistics supposes that, when not completely certain, raters simply guess—a very unrealistic scenario.
The equation for κ is:
where Pr(a) is the relative observed agreement among raters, and Pr(e) is the hypothetical probability of chance agreement, using the observed data to calculate the probabilities of each observer randomly saying each category. If the raters are in complete agreement then κ = 1. If there is no agreement among the raters other than what would be expected by chance (as defined by Pr(e)), κ = 0.
The seminal paper introducing kappa as a new technique was published by Jacob Cohen in the journal Educational and Psychological Measurement in 1960.
A similar statistic, called pi
, was proposed by Scott (1955). Cohen's kappa and Scott's pi
differ in terms of how Pr(e) is calculated.
Note that Cohen's kappa measures agreement between two raters only. For a similar measure of agreement (Fleiss' kappa
) used when there are more than two raters, see Fleiss (1971). The Fleiss kappa, however, is a multi-rater generalization of Scott's pi
statistic, not Cohen's kappa.
Note that there were 20 proposals that were granted by both reader A and reader B, and 15 proposals that were rejected by both readers. Thus, the observed percentage agreement is Pr(a)=(20+15)/50 = 0.70.
To calculate Pr(e) (the probability of random agreement) we note that:
Therefore the probability that both of them would say "Yes" randomly is 0.50*0.60=0.30 and the probability that both of them would say "No" is 0.50*0.40=0.20. Thus the overall probability of random agreement is Pr("e") = 0.3+0.2 = 0.5.
So now applying our formula for Cohen's Kappa we get:
we find that it shows greater similarity between A and B in the second case, compared to the first.
only states how precisely we have measured the magnitude. It makes no claim on how important is the magnitude in a given application or what is considered as high or low agreement.
Statistical significance for kappa is rarely reported, probably because even relatively low values of kappa can nonetheless be significantly different from zero but not of sufficient magnitude to satisfy investigators.
Still, its standard error has been described
and is computed by various computer programs.
If statistical significance is not a useful guide, what magnitude of kappa reflects adequate agreement? Guidelines would be helpful, but factors other than agreement can influence its magnitude, which makes interpretation of a given magnitude problematic. As Sim and Wright noted, two important factors are prevalence (are the codes equiprobable or do their probabilities vary) and bias (are the marginal probabilities for the two observers similar or different). Other things being equal, kappas are higher when codes are equiprobable and distributed similarly by the two observers.
Another factor is the number of codes. As number of codes increases, kappas become higher. Based on a simulation study, Bakeman and colleagues concluded that for fallible observers, values for kappa were lower when codes were fewer. And, in agreement with Sim & Wrights's statement concerning prevalence, kappas were higher when codes were roughly equiprobable. Thus Bakeman et al. concluded that "no one value of kappa can be regarded as universally acceptable." They also provide a computer program that lets users compute values for kappa specifying number of codes, their probability, and observer accuracy. For example, given equiprobable codes and observers who are 85% accurate, value of kappa are .49, .60, .66, and .69 when number of codes is 2, 3, 5, and 10, respectively.
Nonetheless, magnitude guidelines have appeared in the literature. Perhaps the first was Landis and Koch,
who characterized values < 0 as indicating no agreement and 0–.20 as slight, .21–.40 as fair, .41–.60 as moderate, .61–.80 as substantial, and .81–1 as almost perfect agreement. This set of guidelines is however by no means universally accepted; Landis and Koch supplied no evidence to support it, basing it instead on personal opinion. It has been noted that these guidelines may be more harmful than helpful. Fleiss's
equally arbitrary guidelines characterize kappas over .75 as excellent, .40 to .75 as fair to good, and below .40 as poor.
and is especially useful when codes are ordered.
Three matrices are involved, the matrix of observed scores, the matrix of expected scores based on chance agreement, and the weight matrix. Weight matrix cells located on the diagonal (upper-left to bottom-right) represent agreement and thus contain zeros. Off-diagonal cells contain weights indicating the seriousness of that disagreement. Often, cells one off the diagonal are weighted 1, those two off 2, etc.
The equation for weighted κ is:
where k=number of codes and
, 
, and 
are elements in the weight, observed, and expected matrices, respectively. When diagonal cells contain weights of 0 and all off-diagonal cells weights of 1, this formula produces the same value of kappa as the calculation given above.
where
, as usual,
,
k=number of codes,
are the row probabilities, and 
are
the column probabilities.
Others (e.g., Uebersax, 1987) contest the assertion that kappa "takes into account" chance agreement. To do this effectively would require an explicit model of how chance affects rater decisions. The so-called chance adjustment of kappa statistics supposes that, when not completely certain, raters simply guess—a very unrealistic scenario.
Calculation
Cohen's kappa measures the agreement between two raters who each classify N items into C mutually exclusive categories. The first mention of a kappa-like statistic is attributed to Galton (1892), see Smeeton (1985).The equation for κ is:
where Pr(a) is the relative observed agreement among raters, and Pr(e) is the hypothetical probability of chance agreement, using the observed data to calculate the probabilities of each observer randomly saying each category. If the raters are in complete agreement then κ = 1. If there is no agreement among the raters other than what would be expected by chance (as defined by Pr(e)), κ = 0.
The seminal paper introducing kappa as a new technique was published by Jacob Cohen in the journal Educational and Psychological Measurement in 1960.
A similar statistic, called pi
Scott's Pi
Scott's pi is a statistic for measuring inter-rater reliability for nominal data in communication studies. Textual entities are annotated with categories by different annotators, and various measures are used to assess the extent of agreement between the annotators, one of which is Scott's pi...
, was proposed by Scott (1955). Cohen's kappa and Scott's pi
Scott's Pi
Scott's pi is a statistic for measuring inter-rater reliability for nominal data in communication studies. Textual entities are annotated with categories by different annotators, and various measures are used to assess the extent of agreement between the annotators, one of which is Scott's pi...
differ in terms of how Pr(e) is calculated.
Note that Cohen's kappa measures agreement between two raters only. For a similar measure of agreement (Fleiss' kappa
Fleiss' kappa
Fleiss' kappa is a statistical measure for assessing the reliability of agreement between a fixed number of raters when assigning categorical ratings to a number of items or classifying items. This contrasts with other kappas such as Cohen's kappa, which only work when assessing the agreement...
) used when there are more than two raters, see Fleiss (1971). The Fleiss kappa, however, is a multi-rater generalization of Scott's pi
Scott's Pi
Scott's pi is a statistic for measuring inter-rater reliability for nominal data in communication studies. Textual entities are annotated with categories by different annotators, and various measures are used to assess the extent of agreement between the annotators, one of which is Scott's pi...
statistic, not Cohen's kappa.
Example
Suppose that you were analyzing data related to people applying for a grant. Each grant proposal was read by two people and each reader either said "Yes" or "No" to the proposal. Suppose the data were as follows, where rows are reader A and columns are reader B:| B | B | ||
| Yes | No | ||
| A | Yes | 20 | 5 |
| A | No | 10 | 15 |
Note that there were 20 proposals that were granted by both reader A and reader B, and 15 proposals that were rejected by both readers. Thus, the observed percentage agreement is Pr(a)=(20+15)/50 = 0.70.
To calculate Pr(e) (the probability of random agreement) we note that:
- Reader A said "Yes" to 25 applicants and "No" to 25 applicants. Thus reader A said "Yes" 50% of the time.
- Reader B said "Yes" to 30 applicants and "No" to 20 applicants. Thus reader B said "Yes" 60% of the time.
Therefore the probability that both of them would say "Yes" randomly is 0.50*0.60=0.30 and the probability that both of them would say "No" is 0.50*0.40=0.20. Thus the overall probability of random agreement is Pr("e") = 0.3+0.2 = 0.5.
So now applying our formula for Cohen's Kappa we get:
Inconsistent results
One of the problems with Cohen's Kappa is that it does not always produce the expected answer. For instance, in the following two cases there is equal agreement between A and B (60 out of 100 in both cases) so we would expect the relative values of Cohen's Kappa to reflect this. However, calculating Cohen's Kappa for each:| Yes | No | |
| Yes | 45 | 15 |
| No | 25 | 15 |
| Yes | No | |
| Yes | 25 | 35 |
| No | 5 | 35 |
we find that it shows greater similarity between A and B in the second case, compared to the first.
Significance and Magnitude
Statistical significanceStatistical significance
In statistics, a result is called statistically significant if it is unlikely to have occurred by chance. The phrase test of significance was coined by Ronald Fisher....
only states how precisely we have measured the magnitude. It makes no claim on how important is the magnitude in a given application or what is considered as high or low agreement.
Statistical significance for kappa is rarely reported, probably because even relatively low values of kappa can nonetheless be significantly different from zero but not of sufficient magnitude to satisfy investigators.
Still, its standard error has been described
and is computed by various computer programs.
If statistical significance is not a useful guide, what magnitude of kappa reflects adequate agreement? Guidelines would be helpful, but factors other than agreement can influence its magnitude, which makes interpretation of a given magnitude problematic. As Sim and Wright noted, two important factors are prevalence (are the codes equiprobable or do their probabilities vary) and bias (are the marginal probabilities for the two observers similar or different). Other things being equal, kappas are higher when codes are equiprobable and distributed similarly by the two observers.
Another factor is the number of codes. As number of codes increases, kappas become higher. Based on a simulation study, Bakeman and colleagues concluded that for fallible observers, values for kappa were lower when codes were fewer. And, in agreement with Sim & Wrights's statement concerning prevalence, kappas were higher when codes were roughly equiprobable. Thus Bakeman et al. concluded that "no one value of kappa can be regarded as universally acceptable." They also provide a computer program that lets users compute values for kappa specifying number of codes, their probability, and observer accuracy. For example, given equiprobable codes and observers who are 85% accurate, value of kappa are .49, .60, .66, and .69 when number of codes is 2, 3, 5, and 10, respectively.
Nonetheless, magnitude guidelines have appeared in the literature. Perhaps the first was Landis and Koch,
who characterized values < 0 as indicating no agreement and 0–.20 as slight, .21–.40 as fair, .41–.60 as moderate, .61–.80 as substantial, and .81–1 as almost perfect agreement. This set of guidelines is however by no means universally accepted; Landis and Koch supplied no evidence to support it, basing it instead on personal opinion. It has been noted that these guidelines may be more harmful than helpful. Fleiss's
equally arbitrary guidelines characterize kappas over .75 as excellent, .40 to .75 as fair to good, and below .40 as poor.
Weighted Kappa
Weighted kappa lets you count disagreements differentlyand is especially useful when codes are ordered.
Three matrices are involved, the matrix of observed scores, the matrix of expected scores based on chance agreement, and the weight matrix. Weight matrix cells located on the diagonal (upper-left to bottom-right) represent agreement and thus contain zeros. Off-diagonal cells contain weights indicating the seriousness of that disagreement. Often, cells one off the diagonal are weighted 1, those two off 2, etc.
The equation for weighted κ is:
where k=number of codes and
, , and are elements in the weight, observed, and expected matrices, respectively. When diagonal cells contain weights of 0 and all off-diagonal cells weights of 1, this formula produces the same value of kappa as the calculation given above.Kappa Maximum
Kappa assumes its theoretical maximum value of 1 only when both observers distribute codes the same, that is, when corresponding row and column sums are identical. Anything less is less than perfect agreement. Still, the maximum value kappa could achieve given unequal distributions helps interpret the value of kappa actually obtained. The equation for κ maximum is:
where
, as usual,,k=number of codes,
are the row probabilities, and arethe column probabilities.
External links
- its meaning, problems, and several alternatives
- Kappa Statistics: Pros and Cons
- Windows program for kappa, weighted kappa, and kappa maximum
- Java and PHP implementation of weighted Kappa
- AgreeStat: A user-friendly point-and-click Excel VBA program for the statistical analysis of inter-rater reliability data