In many experiments involving finding a suitable replacement plant growth substrate, I have seen ANOVA tukey HSD or Tukey or LSD as post hoc been even if the quantitative factor of increasing rates of compost inclusion (0, 25, 50, 80, 100%) is used. Is this appropriate? what is the argument against using this method and orthogonal polynomials or regression? Examples of which are: 1. http://webs.uvigo.es/jdguez/wp-content/uploads/2012/02/Compost-and-vermicompost-as-nursery-pot-components.pdf
2. http://www.prairieswine.com/pdf/3238.pdf
These articles are available free online
Dear Chaney,
I have read the papers. In general terms, I do not find any argument against the application of ANOVA in those cases. IN my opinion, using the amendment percentag as a categorical facto (predictor) is appropriate. In other fields you can find a wide use of ANOVA in the same way (toxicolgical trails with increasing concentration of chemicals: e.g., 10, 20, 50, 100 (E-6)g/ml in the cell colture medium). Probably I would be more preoccupied by the data (observations) distribution.
Regards
Pier Paolo
Thanks for the reply Pier. My primary concern is that most time that is increasing concentration, there is also a hypothesis as to the effect that the increasing concentration will have relative to the control. So most times there is a trend one expects to see which can be define by polynomials. Given that in the strictness sense these concentrations are not necessarily groups or categorical, but are really quantitative, continuous and most times equally spaced, wouldn't Orthogonal polynomials or regression be better suited for such analysis? Or would it solely hang on the wild card of the objective of the experiments?
I am afraid Chaney but i can't understand you. Why treatment shoud affect the control? In my experience this does not happen. Could possibly provideme more details about you study case? I do not know if i can help you, however.
Regards
Pier Paolo
Pier, my point is why treat proportion increase or rate as a categorical variable when it can be used as a quantitative variable, run the ANOVA, orthogonal polynomial or planned contrast, get differences in mean plus analysis of trend? To me this option makes more use of the data and may make the paper more interesting and robust particularly when one wants to make conclusions about the effect of proportion increase or rate (rank/ordinal) on a particular outcome. To me it is harder to appreciate and be convinced of such an effect when amendment proportions is used as a categorical variable. So this is the angle of the discussion I was hoping to get across for further feedback. It is not that ANOVA is necessarily wrong but the question of how to get the most out of your data to enrich your discussion. Granted may be the question needs to be modified a bit.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics