Esmaeel,

1- Ordinal versus nominal variable?

Kendall’s rank correlation or Kruskal Wallis test

2- Interval versus nominal variable?

Kendall’s rank correlation or one-way analysis of variance

3- Interval versus ordinal variable?

Kendall’s rank correlation or

Spearman’s correlation coefficient

Just check the boxes under Analyze -> Correlation

Spearman’s correlation or Spearman’s rho: Spearman’s rho is a non-parametric statistic and can be used when the data have violated the parametric assumptions (e.g., normality). This test works by first ranking the data and then applying Person’s equation to the ranks.

Kendall’s tau: Kendall’s tau is another non-parametric statistic which should be used rather than Spearman’s coefficient when you have a small data set with large number of tied ranks.

Reference:

Field, A. (2013). Discovering statistics using SPSS (4th ed.). London: Sage.

HTH

Karin

Esmaeel,

1- Ordinal versus nominal variable?

Kendall’s rank correlation or Kruskal Wallis test

2- Interval versus nominal variable?

Kendall’s rank correlation or one-way analysis of variance

3- Interval versus ordinal variable?

Kendall’s rank correlation or

Spearman’s correlation coefficient

Just check the boxes under Analyze -> Correlation

Spearman’s correlation or Spearman’s rho: Spearman’s rho is a non-parametric statistic and can be used when the data have violated the parametric assumptions (e.g., normality). This test works by first ranking the data and then applying Person’s equation to the ranks.

Kendall’s tau: Kendall’s tau is another non-parametric statistic which should be used rather than Spearman’s coefficient when you have a small data set with large number of tied ranks.

Reference:

Field, A. (2013). Discovering statistics using SPSS (4th ed.). London: Sage.

HTH

Karin

FYI- there's a certain amount of (I think) illegitimacy to this typology. I've attached a paper that I think is perhaps the single most comprehensive yet concise description of such a problem.

An excerpt:

"Textbook authors quickly adopted these ideas [the nominal, ordinal, interval, ratio typologies and the corresponding "types" of statistical tests] (e.g. Siegel, 1956, Blalock, 1960), perhaps because they appear to provide simple guidance and protect naive data analysts from errors in applying statistics. Unfortunately, while it seems easy enough to learn to identify the type of scale to which some data might belong, the underlying arguments in terms of transformation classes are subtle and usually not understood by beginning students, and, as we show below, the scale type of data may not be evident at all.

It became common to find charts (often inside the back cover of the text) in which the reader could look up “the appropriate test” based on the number and scale types of the variables at hand." (emphases added)

Thanks Karin and Andrew for your informative responses.

Some more thoughts regarding your question here: https://www.researchgate.net/post/Can_I_use_Pearsons_correlation_coefficient_to_know_the_relation_between_perception_and_gender_age_income

Those who have institutional access to JSTOR can get the final published version of the article Andrew Messing shared here:

https://www.jstor.org/stable/pdf/2684788.pdf

Dmitry Vlasenko thank u