Jump to content

Cronbach's alpha

From Wikipedia, the free encyclopedia
(Redirected from Tau-equivalent reliability)

Cronbach's alpha (Cronbach's ), also known as tau-equivalent reliability () or coefficient alpha (coefficient ), is a reliability coefficient and a measure of the internal consistency of tests and measures.[1][2][3] It was named after the American psychologist Lee Cronbach.

Numerous studies warn against using Cronbach's alpha unconditionally. Statisticians regard reliability coefficients based on structural equation modeling (SEM) or generalizability theory as superior alternatives in many situations.[4][5][6][7][8][9]

History

[edit]

In his initial 1951 publication, Lee Cronbach described the coefficient as Coefficient alpha[1] and included an additional derivation.[10] Coefficient alpha had been used implicitly in previous studies,[11][12][13][14] but his interpretation was thought to be more intuitively attractive relative to previous studies and it became quite popular.[15]

  • In 1967, Melvin Novick and Charles Lewis proved that it was equal to reliability if the true scores[i] of the compared tests or measures vary by a constant, which is independent of the people measured. In this case, the tests or measurements were said to be "essentially tau-equivalent."[16]
  • In 1978, Cronbach asserted that the reason the initial 1951 publication was widely cited was "mostly because [he] put a brand name on a common-place coefficient."[2]: 263 [3] He explained that he had originally planned to name other types of reliability coefficients, such as those used in inter-rater reliability and test-retest reliability, after consecutive Greek letters (i.e., , , etc.), but later changed his mind.
  • Later, in 2004, Cronbach and Richard Shavelson encouraged readers to use generalizability theory rather than . Cronbach opposed the use of the name "Cronbach's alpha" and explicitly denied the existence of studies that had published the general formula of KR-20 before Cronbach's 1951 publication of the same name.[9]

Prerequisites for using Cronbach's alpha

[edit]

To use Cronbach's alpha as a reliability coefficient, the following conditions must be met:[17][18]

  1. The data is normally distributed and linear[ii];
  2. The compared tests or measures are essentially tau-equivalent;
  3. Errors in the measurements are independent.

Formula and calculation

[edit]

Cronbach's alpha is calculated by taking a score from each scale item and correlating it with the total score for each observation. The resulting correlations are then compared with the variance for all individual item scores. Cronbach's alpha is best understood as a function of the number of questions or items in a measure, the average covariance between pairs of items, and the overall variance of the total measured score.[19][8]where:

  • represents the number of items in the measure
  • the variance associated with each item i
  • the variance associated with the total scores

Alternatively, it can be calculated through the following formula:[20]

where:

  • represents the average variance
  • represents the average inter-item covariance.

Common misconceptions

[edit]

[7]

The value of Cronbach's alpha ranges between zero and one

[edit]

By definition, reliability cannot be less than zero and cannot be greater than one. Many textbooks mistakenly equate with reliability and give an inaccurate explanation of its range. can be less than reliability when applied to data that are not essentially tau-equivalent. Suppose that copied the value of as it is, and copied by multiplying the value of by -1.

The covariance matrix between items is as follows, .

Observed covariance matrix

Negative can occur for reasons such as negative discrimination or mistakes in processing reversely scored items.

Unlike , SEM-based reliability coefficients (e.g., ) are always greater than or equal to zero.

This anomaly was first pointed out by Cronbach (1943)[21] to criticize , but Cronbach (1951)[10] did not comment on this problem in his article that otherwise discussed potentially problematic issues related .[9]: 396 [22]

If there is no measurement error, the value of Cronbach's alpha is one.

[edit]

This anomaly also originates from the fact that underestimates reliability.

Suppose that copied the value of as it is, and copied by multiplying the value of by two.

The covariance matrix between items is as follows, .

Observed covariance matrix

For the above data, both and have a value of one.

The above example is presented by Cho and Kim (2015).[7]

A high value of Cronbach's alpha indicates homogeneity between the items

[edit]

Many textbooks refer to as an indicator of homogeneity[23] between items. This misconception stems from the inaccurate explanation of Cronbach (1951)[10] that high values show homogeneity between the items. Homogeneity is a term that is rarely used in modern literature, and related studies interpret the term as referring to uni-dimensionality. Several studies have provided proofs or counterexamples that high values do not indicate uni-dimensionality.[24][7][25][26][27][28] See counterexamples below.

Uni-dimensional data

in the uni-dimensional data above.

Multidimensional data

in the multidimensional data above.

Multidimensional data with extremely high reliability

The above data have , but are multidimensional.

Uni-dimensional data with unacceptably low reliability

The above data have , but are uni-dimensional.

Uni-dimensionality is a prerequisite for . One should check uni-dimensionality before calculating rather than calculating to check uni-dimensionality.[3]

A high value of Cronbach's alpha indicates internal consistency

[edit]

The term "internal consistency" is commonly used in the reliability literature, but its meaning is not clearly defined. The term is sometimes used to refer to a certain kind of reliability (e.g., internal consistency reliability), but it is unclear exactly which reliability coefficients are included here, in addition to . Cronbach (1951)[10] used the term in several senses without an explicit definition. Cho and Kim (2015)[7] showed that is not an indicator of any of these.

Removing items using "alpha if item deleted" always increases reliability

[edit]

Removing an item using "alpha if item deleted"[clarification needed] may result in 'alpha inflation,' where sample-level reliability is reported to be higher than population-level reliability.[29] It may also reduce population-level reliability.[30] The elimination of less-reliable items should be based not only on a statistical basis but also on a theoretical and logical basis. It is also recommended that the whole sample be divided into two and cross-validated.[29]

Ideal reliability level and how to increase reliability

[edit]

Nunnally's recommendations for the level of reliability

[edit]

Nunnally's book[31][32] is often mentioned as the primary source for determining the appropriate level of dependability coefficients. However, his proposals contradict his aims as he suggests that different criteria should be used depending on the goal or stage of the investigation. Regardless of the type of study, whether it is exploratory research, applied research, or scale development research, a criterion of 0.7 is universally employed.[33] He advocated 0.7 as a criterion for the early stages of a study, most studies published in the journal do not fall under that category. Rather than 0.7, Nunnally's applied research criterion of 0.8 is more suited for most empirical studies.[33]

Nunnally's recommendations on the level of reliability
1st edition[31] 2nd & 3rd[32] edition
Early stage of research 0.5 or 0.6 0.7
Applied research 0.8 0.8
When making important decisions 0.95 (minimum 0.9) 0.95 (minimum 0.9)

His recommendation level did not imply a cutoff point. If a criterion means a cutoff point, it is important whether or not it is met, but it is unimportant how much it is over or under. He did not mean that it should be strictly 0.8 when referring to the criteria of 0.8. If the reliability has a value near 0.8 (e.g., 0.78), it can be considered that his recommendation has been met.[34]

Cost to obtain a high level of reliability

[edit]

Nunnally's idea was that there is a cost to increasing reliability, so there is no need to try to obtain maximum reliability in every situation.

Trade-off with validity

[edit]

Measurements with perfect reliability lack validity.[7] For example, a person who takes the test with a reliability of one will either receive a perfect score or a zero score, because if they answer one item correctly or incorrectly, they will answer all other items in the same manner. The phenomenon where validity is sacrificed to increase reliability is known as the attenuation paradox.[35][36]

A high value of reliability can conflict with content validity. To achieve high content validity, each item should comprehensively represent the content to be measured. However, a strategy of repeatedly measuring essentially the same question in different ways is often used solely to increase reliability.[37][38]

Trade-off with efficiency

[edit]

When the other conditions are equal, reliability increases as the number of items increases. However, the increase in the number of items hinders the efficiency of measurements.

Methods to increase reliability

[edit]

Despite the costs associated with increasing reliability discussed above, a high level of reliability may be required. The following methods can be considered to increase reliability.

Before data collection:

  • Eliminate the ambiguity of the measurement item.
  • Do not measure what the respondents do not know.[39]
  • Increase the number of items. However, care should be taken not to excessively inhibit the efficiency of the measurement.
  • Use a scale that is known to be highly reliable.[40]
  • Conduct a pretest - discover in advance the problem of reliability.
  • Exclude or modify items that are different in content or form from other items (e.g., reverse-scored items).

After data collection:

  • Remove the problematic items using "alpha if item deleted". However, this deletion should be accompanied by a theoretical rationale.
  • Use a more accurate reliability coefficient than . For example, is 0.02 larger than on average.[41]

Which reliability coefficient to use

[edit]

is used in an overwhelming proportion. A study estimates that approximately 97% of studies use as a reliability coefficient.[3]

However, simulation studies comparing the accuracy of several reliability coefficients have led to the common result that is an inaccurate reliability coefficient.[42][43][6][44][45]

Methodological studies are critical of the use of . Simplifying and classifying the conclusions of existing studies are as follows.

  1. Conditional use: Use only when certain conditions are met.[3][7][8]
  2. Opposition to use: is inferior and should not be used.[46][5][47][6][4][48]

Alternatives to Cronbach's alpha

[edit]

Existing studies are practically unanimous in that they oppose the widespread practice of using unconditionally for all data. However, different opinions are given on which reliability coefficient should be used instead of .

Different reliability coefficients ranked first in each simulation study[42][43][6][44][45] comparing the accuracy of several reliability coefficients.[7]

The majority opinion is to use structural equation modeling or SEM-based reliability coefficients as an alternative to .[3][7][46][5][47][8][6][48]

However, there is no consensus on which of the several SEM-based reliability coefficients (e.g., uni-dimensional or multidimensional models) is the best to use.

Some people suggest [6] as an alternative, but shows information that is completely different from reliability. is a type of coefficient comparable to Reveille's .[49][6] They do not substitute, but complement reliability.[3]

Among SEM-based reliability coefficients, multidimensional reliability coefficients are rarely used, and the most commonly used is ,[3] also known as composite or congeneric reliability.

Software for SEM-based reliability coefficients

[edit]

General-purpose statistical software such as SPSS and SAS include a function to calculate . Users who don't know the formula have no problem in obtaining the estimates with just a few mouse clicks.

SEM software such as AMOS, LISREL, and MPLUS does not have a function to calculate SEM-based reliability coefficients. Users need to calculate the result by inputting it to the formula. To avoid this inconvenience and possible error, even studies reporting the use of SEM rely on instead of SEM-based reliability coefficients.[3] There are a few alternatives to automatically calculate SEM-based reliability coefficients.

  1. R (free): The psych package[50] calculates various reliability coefficients.
  2. EQS (paid):[51] This SEM software has a function to calculate reliability coefficients.
  3. RelCalc (free):[3] Available with Microsoft Excel. can be obtained without the need for SEM software. Various multidimensional SEM reliability coefficients and various types of can be calculated based on the results of SEM software.

Notes

[edit]
  1. ^ The true score is the difference between the score observed during the test or measurement and the error in that observation. See classical test theory for further information.
  2. ^ This implicitly requires that the data can be ordered, and thus requires that it is not nominal.

References

[edit]
  1. ^ a b Cronbach, Lee J. (1951). "Coefficient alpha and the internal structure of tests". Psychometrika. 16 (3). Springer Science and Business Media LLC: 297–334. doi:10.1007/bf02310555. hdl:10983/2196. S2CID 13820448.
  2. ^ a b Cronbach, L. J. (1978). "Citation Classics" (PDF). Current Contents. 13: 263. Archived (PDF) from the original on 2022-01-20. Retrieved 2021-03-22.
  3. ^ a b c d e f g h i j Cho, Eunseong (2016-07-08). "Making Reliability Reliable". Organizational Research Methods. 19 (4). SAGE Publications: 651–682. doi:10.1177/1094428116656239. ISSN 1094-4281. S2CID 124129255.
  4. ^ a b Sijtsma, K. (2009). "On the use, the misuse, and the very limited usefulness of Cronbach's alpha". Psychometrika. 74 (1): 107–120. doi:10.1007/s11336-008-9101-0. PMC 2792363. PMID 20037639.
  5. ^ a b c Green, S. B.; Yang, Y. (2009). "Commentary on coefficient alpha: A cautionary tale". Psychometrika. 74 (1): 121–135. doi:10.1007/s11336-008-9098-4. S2CID 122718353.
  6. ^ a b c d e f g Revelle, W.; Zinbarg, R. E. (2009). "Coefficients alpha, beta, omega, and the glb: Comments on Sijtsma". Psychometrika. 74 (1): 145–154. doi:10.1007/s11336-008-9102-z. S2CID 5864489.
  7. ^ a b c d e f g h i Cho, E.; Kim, S. (2015). "Cronbach's coefficient alpha: Well known but poorly understood". Organizational Research Methods (2): 207–230. doi:10.1177/1094428114555994. S2CID 124810308.
  8. ^ a b c d Raykov, T.; Marcoulides, G. A. (2017). "Thanks coefficient alpha, we still need you!". Educational and Psychological Measurement. 79 (1): 200–210. doi:10.1177/0013164417725127. PMC 6318747. PMID 30636788.
  9. ^ a b c Cronbach, L. J.; Shavelson, R. J. (2004). "My Current Thoughts on Coefficient Alpha and Successor Procedures". Educational and Psychological Measurement. 64 (3): 391–418. doi:10.1177/0013164404266386. S2CID 51846704.
  10. ^ a b c d Cronbach, L.J. (1951). "Coefficient alpha and the internal structure of tests". Psychometrika. 16 (3): 297–334. doi:10.1007/BF02310555. hdl:10983/2196. S2CID 13820448.
  11. ^ Hoyt, C. (1941). "Test reliability estimated by analysis of variance". Psychometrika. 6 (3): 153–160. doi:10.1007/BF02289270. S2CID 122361318.
  12. ^ Guttman, L. (1945). "A basis for analyzing test-retest reliability". Psychometrika. 10 (4): 255–282. doi:10.1007/BF02288892. PMID 21007983. S2CID 17220260.
  13. ^ Jackson, R. W. B.; Ferguson, G. A. (1941). "Studies on the reliability of tests". University of Toronto Department of Educational Research Bulletin. 12: 132.
  14. ^ Gulliksen, H. (1950). Theory of mental tests. Wiley. doi:10.1037/13240-000.
  15. ^ Cronbach, Lee (1978). "Citation Classics" (PDF). Current Contents. 13 (8). Archived (PDF) from the original on 2022-10-22. Retrieved 2022-10-21.
  16. ^ Novick, M. R.; Lewis, C. (1967). "Coefficient alpha and the reliability of composite measurements". Psychometrika. 32 (1): 1–13. doi:10.1007/BF02289400. PMID 5232569. S2CID 186226312.
  17. ^ Spiliotopoulou, Georgia (2009). "Reliability reconsidered: Cronbach's alpha and paediatric assessment in occupational therapy". Australian Occupational Therapy Journal. 56 (3): 150–155. doi:10.1111/j.1440-1630.2009.00785.x. PMID 20854508. Archived from the original on 2022-10-21. Retrieved 2022-10-21.
  18. ^ Cortina, Jose M. (1993). "What is coefficient alpha? An examination of theory and applications". Journal of Applied Psychology. 78 (1): 98–104. doi:10.1037/0021-9010.78.1.98. ISSN 1939-1854. Archived from the original on 2023-08-13. Retrieved 2022-10-21.
  19. ^ Goforth, Chelsea (November 16, 2015). "Using and Interpreting Cronbach's Alpha - University of Virginia Library Research Data Services + Sciences". University of Virginia Library. Archived from the original on 2022-08-09. Retrieved 2022-09-06.
  20. ^ DATAtab (October 27, 2021). Cronbach's Alpha (Simply explained). YouTube. Event occurs at 4:08. Retrieved 2023-08-01.
  21. ^ Cronbach, L. J. (1943). "On estimates of test reliability". Journal of Educational Psychology. 34 (8): 485–494. doi:10.1037/h0058608.
  22. ^ Waller, Niels; Revelle, William (2023-05-25). "What are the mathematical bounds for coefficient α?". Psychological Methods. doi:10.1037/met0000583. ISSN 1939-1463.
  23. ^ "APA Dictionary of Psychology". dictionary.apa.org. Archived from the original on 2019-07-31. Retrieved 2023-02-20.
  24. ^ Cortina, J. M. (1993). "What is coefficient alpha? An examination of theory and applications". Journal of Applied Psychology. 78 (1): 98–104. doi:10.1037/0021-9010.78.1.98.
  25. ^ Green, S. B.; Lissitz, R. W.; Mulaik, S. A. (1977). "Limitations of coefficient alpha as an Index of test unidimensionality". Educational and Psychological Measurement. 37 (4): 827–838. doi:10.1177/001316447703700403. S2CID 122986180.
  26. ^ McDonald, R. P. (1981). "The dimensionality of tests and items". The British Journal of Mathematical and Statistical Psychology. 34 (1): 100–117. doi:10.1111/j.2044-8317.1981.tb00621.x.
  27. ^ Schmitt, N. (1996). "Uses and abuses of coefficient alpha". Psychological Assessment. 8 (4): 350–3. doi:10.1037/1040-3590.8.4.350.
  28. ^ Ten Berge, J. M. F.; Sočan, G. (2004). "The greatest lower bound to the reliability of a test and the hypothesis of unidimensionality". Psychometrika. 69 (4): 613–625. doi:10.1007/BF02289858. S2CID 122674001.
  29. ^ a b Kopalle, P. K.; Lehmann, D. R. (1997). "Alpha inflation? The impact of eliminating scale items on Cronbach's alpha". Organizational Behavior and Human Decision Processes. 70 (3): 189–197. doi:10.1006/obhd.1997.2702.
  30. ^ Raykov, T. (2007). "Reliability if deleted, not 'alpha if deleted': Evaluation of scale reliability following component deletion". The British Journal of Mathematical and Statistical Psychology. 60 (2): 201–216. doi:10.1348/000711006X115954. PMID 17971267.
  31. ^ a b Nunnally, J. C. (1967). Psychometric theory. McGraw-Hill. ISBN 0-07-047465-6. OCLC 926852171.
  32. ^ a b Nunnally, J. C.; Bernstein, I. H. (1994). Psychometric theory (3rd ed.). McGraw-Hill. ISBN 0-07-047849-X. OCLC 28221417.
  33. ^ a b Lance, C. E.; Butts, M. M.; Michels, L. C. (2006). "What did they really say?". Organizational Research Methods. 9 (2): 202–220. doi:10.1177/1094428105284919. S2CID 144195175.
  34. ^ Cho, E. (2020). "A comprehensive review of so-called Cronbach's alpha". Journal of Product Research. 38 (1): 9–20.
  35. ^ Loevinger, J. (1954). "The attenuation paradox in test theory". Psychological Bulletin. 51 (5): 493–504. doi:10.1002/j.2333-8504.1954.tb00485.x. PMID 13204488.
  36. ^ Humphreys, L. (1956). "The normal curve and the attenuation paradox in test theory". Psychological Bulletin. 53 (6): 472–6. doi:10.1037/h0041091. PMID 13370692.
  37. ^ Boyle, G. J. (1991). "Does item homogeneity indicate internal consistency or item redundancy in psychometric scales?". Personality and Individual Differences. 12 (3): 291–4. doi:10.1016/0191-8869(91)90115-R.
  38. ^ Streiner, D. L. (2003). "Starting at the beginning: An introduction to coefficient alpha and internal consistency". Journal of Personality Assessment. 80 (1): 99–103. doi:10.1207/S15327752JPA8001_18. PMID 12584072. S2CID 3679277.
  39. ^ Beatty, P.; Herrmann, D.; Puskar, C.; Kerwin, J. (July 1998). ""Don't know" responses in surveys: is what I know what you want to know and do I want you to know it?". Memory (Hove, England). 6 (4): 407–426. doi:10.1080/741942605. ISSN 0965-8211. PMID 9829099. Archived from the original on 2023-02-20. Retrieved 2023-02-20.
  40. ^ Lee, H. (2017). Research Methodology (2nd ed.), Hakhyunsa.
  41. ^ Peterson, R. A.; Kim, Y. (2013). "On the relationship between coefficient alpha and composite reliability". Journal of Applied Psychology. 98 (1): 194–8. doi:10.1037/a0030767. PMID 23127213.
  42. ^ a b Kamata, A., Turhan, A., & Darandari, E. (2003). Estimating reliability for multidimensional composite scale scores. Annual Meeting of American Educational Research Association, Chicago, April 2003, April, 1–27.
  43. ^ a b Osburn, H. G. (2000). "Coefficient alpha and related internal consistency reliability coefficients". Psychological Methods. 5 (3): 343–355. doi:10.1037/1082-989X.5.3.343. PMID 11004872.
  44. ^ a b Tang, W., & Cui, Y. (2012). A simulation study for comparing three lower bounds to reliability. Paper Presented on April 17, 2012 at the AERA Division D: Measurement and Research Methodology, Section 1: Educational Measurement, Psychometrics, and Assessment, 1–25.
  45. ^ a b van der Ark, L. A.; van der Palm, D. W.; Sijtsma, K. (2011). "A latent class approach to estimating test-score reliability". Applied Psychological Measurement. 35 (5): 380–392. doi:10.1177/0146621610392911. S2CID 41739445. Archived from the original on 2023-08-13. Retrieved 2023-06-04.
  46. ^ a b Dunn, T. J.; Baguley, T.; Brunsden, V. (2014). "From alpha to omega: A practical solution to the pervasive problem of internal consistency estimation" (PDF). British Journal of Psychology. 105 (3): 399–412. doi:10.1111/bjop.12046. PMID 24844115. Archived (PDF) from the original on 2023-03-24. Retrieved 2023-06-04.
  47. ^ a b Peters, G. Y. (2014). "The alpha and the omega of scale reliability and validity comprehensive assessment of scale quality". The European Health Psychologist. 1 (2): 56–69.
  48. ^ a b Yang, Y., & Green, S. B.Yanyun Yang; Green, Samuel B. (2011). "Coefficient alpha: A reliability coefficient for the 21st century?". Journal of Psychoeducational Assessment. 29 (4): 377–392. doi:10.1177/0734282911406668. S2CID 119926199.
  49. ^ Revelle, W. (1979). "Hierarchical cluster analysis and the internal structure of tests". Multivariate Behavioral Research. 14 (1): 57–74. doi:10.1207/s15327906mbr1401_4. PMID 26766619.
  50. ^ Revelle, William (7 January 2017). "An overview of the psych package" (PDF). Archived (PDF) from the original on 27 August 2020. Retrieved 23 April 2020.
  51. ^ "Multivariate Software, Inc". www.mvsoft.com. Archived from the original on 2001-05-21.
[edit]