I have an lme model in R with three main effects (sediment size, concentration and algae concentration). After doing the necessary data processing I arrived at my minimal model which comprises of two different two-way interactions (sediment size:sediment concentration and algae concentration: sediment size). I used glht to conduct pairwise comparisons for these two two-way interaction and managed to also get the alphabets for significant differences through the cld function. With the two different two-way interactions I would essentially have two sets of significant differences, so I need suggestions as to how best to display this graphically. I have them in a table now and I have been told that it is too complicated for readers.
If I understand you correctly, then you have continuous variables that interact (as well as have main effects) on a outcome variable? Depending on the number of cases, one approach is to group the data by independent variable - you could do this by recoding, e.g., sediment size into a binary variable (code 1 as > mean+1SD, code 0 as < mean-1SD). For the purposes of graphing, you'd then think of the independent variables as factors/conditions rather continuous measures. On the y axis would be your outcome variable, and you'd put e.g. sediment size (high/low) on the x axis, with points/bar heights on the graph according to outcome variable for the sediment concentration (high/low). This would give you a graph with four points or bars. I've done something (sort of) similar in: Article Impacts of adaptation and responsibility framings on attitud...
Hi Stuart, thanks for your suggestion. Just to answer you question, of my three main effects I have two continuous variables (sediment and algae concentration) and one categorical (sediment size) and my dependent variable is ingestion rate. For both sediment size and algae concentration I have three levels but for sediment concentration I have five levels. It's quite a number of combinations. For the article, are you referring to the table 1 with the linear regression results?
Hi Genevieve - yes, that is a fair number of combinations. It might be that you need to construct separate graphs, e.g. with sediment size on x axis (separate bars), y axis ingestion, and height of separate bars according to another variable, e.g. algae concentration? Then repeat for another combination. I guess it depends a lot on which interactions are most important or hypothesised as to how you set this up. In our paper, Table 1 gives the results in numerical form; Fig 1 graphically (in the latter, lines used to show interaction pattern, though you might be better with bars).
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping
- Maps