I am looking at the impact of a fungus on a plant over time by measuring key plant parameters (height, number of flowers, seeds, leaves etc) three times over the growing season. I have 12 replicates for each treatment and three treatments - two rust strains and a control.

What stats test would be most useful to determine any differences between treatments but also look take in consideration the impact of time? A mixed effect model would group time together to give the overall trend, but what if the effect was more noticeable at one of the time points for example? Is it ok to compare the time points separately or is there a more suitable statistical test that I should use?


Given that you only have three time points and 36 measurements per time point (i.e., 12 per treatment), you could use a marginal multivariate regression model to assess differences over time. Using the nlme package in R, you could try something along these lines:


gls(plant_height ~ time * treatment, data = <your_data>,
    correlation = corSymm(form = ~ 1 | plant_id),
    weights = varIdent(form = ~ 1 | time))

where time should be coded as a factor in your dataset. Then using the anova() function you could test for the interaction term between treatment and time that would tell you if there is a statistical indication for treatment differences over time.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.