I have the results from an experiment . The groups are
1. Saline
2. Treatment
3. Saline + X inhibitor
4. Treatment + X inhibitor
In previous experiments, treatment had a remarkable effect. My hypothesis was that treatment acted by influencing X and inhibiting X will make Saline and treatment group same. The experiment seems to have proved my hypothesis.
My comparison groups are 1 vs 2 or 3 vs 4. I am not at all concerned about whether the mean of 3 or 4 are same as that of 1 or 2.
Results are shown below
To compare 1 vs 2 and 3 vs 4 I used two Welch’s t tests instead of ANOVA. Is the Welchs test robust in my cases (my data is normal but may not same variance)
Thanks
Dear Sethi
It depends on how you present your results and how can advocate your answers.
Comparing any unrelated(independent) two groups should be developed using two-independent T test if the sample size is more than 30. Otherwise, paired T-Test should be developed.
Regards,
Zuhair
First, fix the typo in your y-axis label. ;)
You can argue that you are making planned comparisons, but reviewers may want you to first do the omnibus ANOVA and then conduct the two planned comparisons to control for family wise error. Heterogeneity of variance may be an issue, however.
I presume that all conditions were between subject. Good luck!
I agree with Michael E Young: you should do a 2x2 ANOVA, and in particular look at the test for the interaction effect. Depending on factors like sample size, heterogeneity of variance, etc. it appears that you have an interaction effect, where salinity matters less when the inhibitor is present.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA