I have an experiment having 39 different ber fruit varieties with 3 replications of each. I want one single LSD value for each biochemical such as vitamin, phenol, acidity and TSS of all ber varieties separately. But Spss shows mean difference and standard error of each single variety with other varieties. But I want single LSD value for each biochemical as shown in file attached.
Hello Jayant,
If I understand your question, you're concerned that the Fisher LSD method uses unique standard errors for each possible pairing of varieties. There's a simple fix, and that is, don't use the Fisher LSD! For one thing, it offers no strict control over aggregate Type I error risk (and, with 39 varieties to be compared on three attributes, that's a _lot_ of tests!). Both the Tukey HSD and the Scheffe method invoke a common standard error (so that there would be a single significance criterion in force within a single attribute). The Scheffe is overly conservative, however, so you might opt for the Tukey HSD instead (or, apply the Scheffe at a higher alpha level than normal, e.g., 10% risk instead of 5% or 1% risk).
If you absolutely, positively, must use repeated t-tests (aka, Fisher LSD) without protection, then use the pooled, within-groups MS as the error term instead of the individual variety's variance.
Good luck with your work!
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS