Skip to main content
Home
About repository
Contact
EN
HR
Repository of the University of Rijeka
UNIRI Repository
Browse
By Author
By Year
By Organizational unit
By Scientific field
By Object type
By repositories
All documents
Advanced search
Upload
Search Term
Download
PDF 3.52 MB
Professional paper - Review paper
0
0
Characteristics of categorical data analysis
Zuvic-Butorac, Marta (2006)
Cite this item:
https://urn.nsk.hr/urn:nbn:hr:193:265096
Metadata
Language
Croatian
Title (English)
Characteristics of categorical data analysis
Author
Zuvic-Butorac, Marta
Abstract (Croatian)
This work presents characteristics of categorical data, their presentation and possible models of statistical analysis. There are two types of categorical data; nominal, where categories are equally valued (i.e. we can measure them only in terms of whether individual items belong to some distinctively different categories) and ordinal, where categories allow to rank order on some scale of measurement. Categorical data are non-numeric by nature, but could be numerically presented and analyzed anyway. Although the information value of categorical data is well below the respective information value of numerical, practically there's no study where their analysis wouldn't be of importance. Numerically, the categorical data can be presented via frequencies (absolute number of items belonging to the category) or their proportion (percentage) in the sample. Adequate graphical presentation goes with pie charts or percentage stacked bars charts. The statistical analysis of categorical data most often is done on contingency tables. The type of the analysis depends on the relation between samples from which the data are drowning. If the samples are independent, the analysis would be performed using difference of proportion test, chi2 test or Fisher exact test. The limitations and suitability for application of the three is discussed. If the samples are dependent, the choice goes to McNemar chi2 test (2 samples) or Cochrane's Q test 8more than 2 samples). The conclusions from the aforementioned analyses could be drowned only in terms of significant or nonsignificant relations between the rows and columns in the contingency tables. In the need to measure the level of relations between categorical data, two types of measures are defined: relative risk and odds ratio, which can both be calculated only in 2x2 contingency tables. Relative risk is a ratio of two proportions (suitable only in prospective studies), whether odds ratio measures ratio between odds in two groups (could be calculated both in prospective and retrospective studies). All the aforementioned analyses are well documented with calculations on data collected in biomedical studies.
Keywords (Croatian)
categorical data
frequency
proportion
contigency table
chi-square test Fisher exact test
McNemar test
Cochrane Q test
odds ratio
Publication type
professional paper - review paper
Publication status
published
Peer review
peer review - domestic
Journal title
Acta medica Croatica
Numbering
2006, Vol. 60, No. 1, pp 63-79
ISSN
1330-0164
e-ISSN
1848-8897
Date
publication: 2006.
Article URL
https://www.ncbi.nlm.nih.gov/pubmed/16526308
Scientific field
NATURAL SCIENCES
Mathematics
Probability Theory and Statistics
Institution
University of Rijeka
(Department of Biotechnology)
URN:NBN
https://urn.nsk.hr/urn:nbn:hr:193:265096
Repository
https://repository.biotech.uniri.hr