For example... in a completely randomized design (DIC), 3 treatments with 3 replicates... it would be 8 degrees of freedom... if i have 3 treaments with 4 replicates, it would be 11 degrees of freedom...
Is there a minimum value to obtain an adequate statistical analysis?
Would be some value above the 10?
im asking about the degree of freedom in the experimental design...
How many replicates would be necessary? No matter what the kind of design, block.. latin square... factorial design... split plot...
The more replicates, the more degree of freedom. Is there some reference or some statistical rule for a minimum value?
It has been found that degrees of freedom in any design improves the reality of results obtained, and the minimum of d.f is 1, when we use Chi square of one degree of freedom to compare two means under question. Regards.
Hi Peterson,
The required number of replicates should be based on the minimum amount of ten degrees of freedom for the residual. This information is consolidated in some works.
GOMES, Frederico Pimentel. Curso de Estatística Experimental. 11. ed. Piracicaba-sp: Revista e Ampliada, 1995.
NESI, Cristiano Nunes et al. Número mínimo de repetições em experimentos de competição em híbridos de milho. Revista de Ciências Agroveterinárias, Lages, v. 9, n. 1, p.74-81, 19 fev. 2010.
Hugo Leandro dos Santos , I am curious why you disagree with the responses above, since you presumably read them prior to responding. If a medical research has a control group and a treatment group, and the outcome is whether the person recovers, and suppose they have a thousand people in each condition, they might use a chi-squared test (perhaps with some adjustments) and this has 1 degree of freedom in this situation. Are you saying this would not be right?
Peterson Emmanuel Guimarães Paixão, I may be wrong, but the examples you gave suggest to me that you are asking about one-way ANOVA specifically. Is that right?
Bruce Weaver , his question "Is there any minimum value for degrees of freedom in an experimental design?" seems more general. If he meant ANOVA with an F distribution, there would be multiple dfs so he would have said which one (or both).
Hello Daniel Wright. He also wrote this:
For example... in a completely randomized design (DIC), 3 treatments with 3 replicates... it would be 8 degrees of freedom... if i have 3 treaments with 4 replicates, it would be 11 degrees of freedom...
I believe that replicates means subjects in our (psychology) vernacular. For the first example, then, 3 treatments with 3 replicates means 3 groups with 3 Ss per group. For a one-way ANOVA model, that would give dfTotal = 9 - 1 = 8. For the other example that was given, 3 groups x 4 Ss/group = 12 Ss in total. So dfTotal = 12 - 1 = 11. This is why I speculate that the question is really about one-way ANOVA. ;-)
Could be. I thought he was asking experiments in general, and then was just giving an example of one type of experimental design. Maybe he is even just asking about power, or something like that. Given it not clear, at least to me, I won't comment anymore.
- Badges
- Science topic