I think this is a bit silly, but the people have spoken.
I view this as a marketing ploy, actually -- if users really want plots, they will chose a package that has them, and they might get sucked into some suboptimal method that way.
I'd rather empower people to learn ggplot, but 🤷
I guess it could be a good "sanity check" for large/small effect sizes in simple models
Plotting data
So since "radEmu is the well-defined and sensible version of a CLR linear model", perhaps taxa on the y-axis, CLR-transformed data on x-axis, colors for main constrast... it should be a ggplot, so people can add to it.
Half the lowest value per sample as pseudocount?
Plotting estimates (main contrast)
Our usual ordered plot that we love (x = taxa ordered by effect size, y = effect size + CI), made easy.
Plotting estimates (all estimates)
A heatmap. Taxa on y axis, covariates on x axis.
I think this is a bit silly, but the people have spoken.
I view this as a marketing ploy, actually -- if users really want plots, they will chose a package that has them, and they might get sucked into some suboptimal method that way.
I'd rather empower people to learn ggplot, but 🤷
I guess it could be a good "sanity check" for large/small effect sizes in simple models
Plotting data
So since "radEmu is the well-defined and sensible version of a CLR linear model", perhaps taxa on the y-axis, CLR-transformed data on x-axis, colors for main constrast... it should be a ggplot, so people can add to it.
Half the lowest value per sample as pseudocount?
Plotting estimates (main contrast)
Our usual ordered plot that we love (x = taxa ordered by effect size, y = effect size + CI), made easy.
Plotting estimates (all estimates)
A heatmap. Taxa on y axis, covariates on x axis.