I have a sample size of 6 for my Spearman rank correlation.
However, Fowler et al., 1998 (Practical statistics for field biology. 2nd Edition. England, UK: John Wiley and Sons. pp 259) states that a sample size of at least 7 is needed to perform the analysis.
Can I still use my results?
Thank you!
Hello Rebecca,
You can compute a correlation coefficient with as few as 2 cases. However, the result is guaranteed to be either +/- 1 or undefined (due to zero variance). So, the smallest sample size that doesn't yield such nonsensical values is 3 (and even this restricts values to +/- 1, or +/- 0.5, or undefined).
You can, therefore, compute a Spearman correlation with an n of 7. It will not offer much statistical power for testing the null hypothesis that rho = 0, however. The minimum value needed for a directional hypothesis test to be significant at the .05 level is .714 (.893 at the .01 level).
Good luck with your work.
To know more about sample size for Spearman rank correlation, I suggest you to check the following links
https://stats.stackexchange.com/questions/415131/minimum-sample-size-for-spearmans-correlation-and-kendalls-tau-b
Article Sample size requirements for estimating Pearson, Kendall and...
Hello Rebecca,
You can compute a correlation coefficient with as few as 2 cases. However, the result is guaranteed to be either +/- 1 or undefined (due to zero variance). So, the smallest sample size that doesn't yield such nonsensical values is 3 (and even this restricts values to +/- 1, or +/- 0.5, or undefined).
You can, therefore, compute a Spearman correlation with an n of 7. It will not offer much statistical power for testing the null hypothesis that rho = 0, however. The minimum value needed for a directional hypothesis test to be significant at the .05 level is .714 (.893 at the .01 level).
Good luck with your work.
David Morse , can you re-phrase your final sentence? I think you are saying something about some power amount if this minimum effect to detect is rho = .714 and n=7. The way phrased it might be interpreted in different ways.
Daniel Wright , the sentence is correct, however, let me add a few clarifying words: The minimum observed (Spearman rho) value needed to reject the null hypothesis (that population rho = 0) is .714 (directional test) at the .05 level, or .893 at the .01 level (with an N of 7).
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size