### Topics

- As in other areas of the life, there isn't a unique solution to your problem. To choose correctly you must think about what kind of error do you consider more important to prevent: Type I error (considering significant a difference that actually isn't significant) or type II error (considering non significant a difference that actually is singificant). If you chooses to prevent Type I error, you should use Tukey test, otherwise, use Duncan test.
- You need a control group as well so the design is 0.0, 2.5, 4.5 and 6.5 mg/ml - planned comparisons are better than post-hoc tests since they are less conservative - if your data are parametric you can use t--tests with bonferroni correction - comparing 0 with 2.5, 0 with 4.5, 0 with 6.5 - can also do 2.5 v 4.5 and 6.5 and 4.5 v 6.5 - this will give you 6 comparisons on one data set, so you might also consider increasing acceptance level from 0.05 to 0.01 - there are other options for your pair-wise comparisons but this will be the simplest and easiest to do
- Thank you so much for the help, your comments have definitely provided clarity.( I do have a control group - I meant to include this information in the question.)
- You need a trend analysis and Quadratic regression will show the optimum level of the treatment if you are rightly choosing the levels

- Duncan's multiple range test is appropriate
- Tukey’s Studentized Range (HSD) test is appropriate and very eligible.
- Duncan's multiple range test
- the best test isDuncan's multiple range
- Thank you to everyone for your help, it is much appreciated
- I think that my comments comes late anyhow you can use Duncan or Tukey tests as they are multiple range tests to compare more than two means and also you can use t-test despite this test is more suitable to compare only two means
- T-test is easy and suitable to your case
- You can use Tukey and Dencan test as you like to compare your treatments.
- Your question has been aswred as far as posthoc is concerned. But I think you could gain more insight with a regression analysis, especially if your data has breath enough to be meaningfully mined with such an approach. At least we see that you are studying the response at 3 different levels. You could check if the relationship is linear of nonlinear, leading you then as part of your conclusion to check if there is room to change the dosing levels- either to increase or not, for a nicer contribution to knowledge.
- You can use either Tukey or Duncan post-hoc tests for evaluation which dosages effects were different, but tukey test also compare each group with other group (2 by 2) like a t-test, and it let you to know what is the exact different between groups.
- In one-way anova the two tests are equally eligible.
- You can use either Tukey or Duncan post-hoc tests for evaluation......
- Tukey or Duncan or LSD up to you
- I use to check if the data are really parametric (normality). Most of the cases they are not.

If they are parametric you could use a 2-ways anova with the treatment and boar. It is well known the variabiliry related to individual differences. So, if you control and monitorice this individual variability you can consider the real effect of the treatment, and the interaction between treatment and boars.

Good luck with your experiments, antioxidant treatment is a amazing topic!!!

Joaquín - agree, should researcher consider in effect of boar. if the researcher use RCBD randomized complete block design (TWO-way ANOVA) and block male. It could be test interaction that it shown significant or not
- Tukey's range test will be a better option ..
- Tukey's range test gives you more accurate differetiation among different groups as compared to duncan's multiple range test
- Tukeys is what I used
- Any of the two mean comparison procedure will work well except you have another group where you have 0 mg/ml of ascorbic acid (control). if there is a control treatment (0 mg/ml of ascorbic acid) then Dunnett will be appropriate
- you can use more than one method and then make a simple comparison among , it depends on the standard deviation values, some of that tests are hard. you can use LSD, Tukey, and Duncan
- In my experience, protected LSDs are a good way of assessing the extent when looking at doses (e.g. your 2.5 - 6.5 mg/ml) where you expect your treatments may not be as distinct categories as things like variety or boar. Since when using multiple comparison tests like Tukey's you make the assumption/expect that each treatment could be quite distinct from another. The factorial design mentioned by Joaquín looking at your treatments and boar is in my opinion a very useful design that could provide a lot of information, provided you have have enough replicates should you find interactions.
- None of the so-called mean separation tests are as robust as orthogonal contrasts. In this case, because you have equally spaced dilutions, you can use quadratic orthogonal contrast.
- If you have groups of the same size (= n) Tukey´s t test is probably the best option. If your groups have different number of observations, Bonferroni´s test is more appropriated.
- Turkey's range test
- I would say to use bonferroni post hoc test as it is the simplest and conservative method. Also if you conduct the ANOVA using Mixed Model then use Least squares Means differences
- Tukey's range test or duncan's multiple range test
- For equally spaced treatments as in this case, orthogonal contrasts is more appropriate so as to see clearly the trends among the treatments. If 2.5mg/ml is the standard, then treatment pair comparison using contrast statement will give a better result.
- As two previous commentators have stated, orthogonal contrasts are much more robust than t tests to determine if means are different.
- Tukey's test with orthogonal contrast is more appropriate

## All Answers (33)