Dear Lee,

Five responses is a bit low for correlation. Because your data type is ordinal you would need to use a non-parametric rank correlation, such as Spearman. However, the most appropriate test is probably Chi-squared, but this will detect an association rather than a correlation - a weaker concept of some kind of relationship. You may need to group together the higher number of horses owned in order to satisfy the validity of the chi-squared test - see our guide: https://www.researchgate.net/publication/301230965_Chi-Squared_Test_for_Two-Way_Tables_Revised

Data Chi-Squared Test for Two-Way Tables (Revised)

Hello Lee

Please check the following links which may be useful to answer your question!

https://www.researchgate.net/post/How_can_we_do_analysis_of_the_Likert_Scale_in_SPSS

https://www.researchgate.net/post/What_statistical_analysis_should_I_use_for_Likert-Scale_data

https://www.researchgate.net/post/How_to_Best_Analyze_2_Groups_Using_Likert_Scales_in_SPSS

https://www.researchgate.net/post/What_statistical_analysis_should_I_use_for_Likert-Scale_data2

https://www.researchgate.net/post/How_best_can_you_analyse_likert_scale_data_using_SPSS

https://www.researchgate.net/post/What_statistics_does_one_use_for_a_Likert_scale

https://www.researchgate.net/post/Appropriate_statistical_test_for_Likert_items_and_scales_to_find_out_if_there_is_any_significant_difference_in_the_responses_of_3_different_groups2

https://www.google.com.sa/url?sa=t&source=web&rct=j&url=https://students.shu.ac.uk/lits/it/documents/pdf/questionnaire_analysis_using_spss.pdf&ved=0ahUKEwjdkrTWuI3SAhUBBBoKHUCxCa4QFggaMAE&usg=AFQjCNEZsaGGdA-Il-LI07XTumlyTgUhdw

https://www.google.com.sa/url?sa=t&source=web&rct=j&url=https://students.shu.ac.uk/lits/it/documents/pdf/analysing_data_using_spss.pdf&ved=0ahUKEwjdkrTWuI3SAhUBBBoKHUCxCa4QFggcMAI&usg=AFQjCNEheDT4eyWUkl8yL7NrdGuO2M9z4Q

https://www.google.com.sa/url?sa=t&source=web&rct=j&url=http://www.jgme.org/doi/pdf/10.4300/JGME-5-4-18&ved=0ahUKEwiT2Omnuo3SAhUDPRQKHRvHArw4ChAWCDIwCQ&usg=AFQjCNGk4cXfLeQB3XyUh-gd7ZsbmKcWiw

I hope this could help you

Best wishes

Mansour

Thanks so much for you answer, Peter! I am using Chi-squared for most of my questions. If I wanted to use Spearman's I somehow need to rank the data, don't I? I attached you a screenshot - I assigned numbers to the Likert scales. First column shows you the number of horses and second column risk tolerance (e.g. strongly agree = 5).

Thank you Mansour!

Dear Lee,

I think the best analysis for the questions in your screenshot would be Chi-squared, not Spearman's rank. As there are low frequencies of 1 and 2 on your Likert response scale (which is quite common) I would recommend computing a new variable where you combine 1 and 2 with 3 to make the new value "1 to 3". I would also recommend combining values on the no. horses scale into intervals in a geometric way, e.g. 1, 2-3, 4-7, 8-15, 16+. You should then be able to satisfy the validity restrictions of the Chi-squared test (expected frequencies > 5 in 80% of the cells).

Thanks for your help, Peter!