Hi Altair,

If I understood correctly, you have entered the data in wide format (one row per observationn, one colum per variable), so no problems there. If you can screenshot part of the data frame and part of the output that you find lengthy, perhaps I can help more.

On the other, unrelated note I have one question. Purely out of academic interest do you expect to see decrease in percived simptoms of DOMS in groups that stretched?

Take care.

Ivan

Hm, from what I can tell it looks as if you have entered all the variables in one ANOVA test. That's why you got that uinteligable output.

You should do ANOVA for every muscle separately. Factors are time and group, outcome is VAS/ROM. Since you have to outcomes (dependant variables - DVs) MANOVA could be more appropriate in order to detect subtler differences since VAS and ROM could be correlated (especially if you have measured active ROM). It also worth noting that MANOVA will cut your ANOVA's by half which could somewhat help you avoid type I error, although if both outcomes are significant and you need to do both ANOVA's per muscle, you should control for multiple testing issue.

If you are confused about the analysis I can recommend reading

Discovering Statistics Using IBM SPSS Statistics by Andy Field. Very fun and light read with many helpful hints and bunch of traning datasets.

Also, three other things:

1. What's your sample size.

2. ROM was measured by hand goniometer, electric one or by kinematic methods.

3. VAS used is 10-points or 100-points.

Hm, 9 participants (3 per group) is *very* low. Both hand goniometer and VAS are somewhat imprecise. They are excellent clinical tools for quick assessment of a patient but for measuring and general inference, not so much. On top of all that, the effect of stretching on the perceived discomfort of DOMS is probably very small1.

Given all these, I doubt you will find the difference between groups and even if you did so by pure chance, you definitely could not trust results with such a small sample. I know this work is for your undergrad thesis but if it is in any way possible, you should measure more people.

How much more people you need to detect a difference depends how far apart are the means of the group, how big is the variance around the means, and how large is the effect size. We know, from the above-cited paper, that effect size is small, variances are, given the measuring tools and general unpredictability of biology, solidly large so in order to detect statistically significant differences in groups, you probably need a lot of people. Is that stat. significant difference relevant is an entirely different question. You should probably talk about this with you thesis coordinator or mentor.

All of the above, although important, still did not answer your questions, so:

1. Left and right muscle should be differently noted. You *could* double the sample sizes by erasing left or right, but then you observations are not independent and that breaks some assumptions about statistical testing.

2. Unless you are doing MANOVA, every outcome is a different ANOVA with a different output. In papers (and thesis') you create a table with a few key statistics from the outputs and collate them all into that table (or tables).

3. Shapiro-Wilk W test: Usually yes, but you can also look at qq plots or histograms and evaluate them visually. W test has a greater chance of type II error on small samples, however, I do not think that ROM will be distributed normally anyhow because of the ceiling effect.

4. Whatever looks more informative at a glance. Bar charts would be my guess in this instance.

5. Every outcome is a different ANOVA.

6. As stated, you should not combine data from left and right muscle because those measures are not independent.

I hope I helped but I would recommend finding someone who can assist you in person. It will be much more helpful than online help.

Take care!

1Herbert RD, de Noronha M, Kamper SJ. Stretching to prevent or reduce muscle soreness after exercise. Cochrane Database of Systematic Reviews 2011, Issue 7. Art. No.: CD004577. DOI:10.1002/14651858.CD004577.pub3.

1. F-ratio, p-value, partial eta squared, if there are any statistically significant findings and perhaps degrees of freedom.

Yes, however, I must point out that given low sample and general purpose of the analysis you would have been better off if you could have measured one muscle on more subjects. Tables and plots for any findings you wish to summarize, however, if you have measured, analysed and then dropped any variable out of the report it should be accounted for. I'm including an example of a table in the attachment. BA for within effects, ES for between effects and BA x PP for interaction.

2. Definitely different plots for VAS. Y-axis is 1-10 for VAS and ~ 0-145 for knee flexion

3. Simple main effects are useful however some authors caution against it and recommend using simple contrasts instead. But I believe that, given the small sample size and small effect size that you probably won't need to worry about that.

Hi, Altair

To understand what are you trying to demonstrate, I‘d like to know the title of your investigation and your hypothesis.