I am analyzing an experiment where I grew two genotypes (native and invasive) of the same plant in one pot and evaluated growth over time. Half of the pots had two plants of the same genotype (native-native) and the other half with one of each genotype (native-invasive). These pots were then either placed in a herbivory or no herbivory treatment. Each plant was then harvested, dried, and weighed separately. I would like to compare the growth of each plant in the two genotype (competition) treatments to each other and evaluate if herbivory affected these growth patterns. Since two plants were grown in one pot, do I need to somehow nest plants in pot?

Research question: does herbivory differentially affect native genotypes grown with other natives versus native plants grown with invasive genotypes?


1 Answer 1


The way your question is phrased, it seems that the plant is actually the sample, that is to say, what the damage is being measured on.

In order for the pot to be the sample, and if you want to compare the growth of the 2 genotypes, I think I would calculate the weight difference between the 2 plants of 1 pot and make this the response variable. Then you would have one variable genotype of the “potmate” with 2 levels matching/mismatching, and one variable herbivory with 2 levels = yes/no. In R, the model would look something like this : weight difference~herbivory*genotype of the “potmate”. Here there are no random factors, which you referred to as “nested” I guess.

However since you have measured damage on each plant, you could consider one plant as a sample and the damage to one given plant (measured as weight at the end of the experiment as I understand it) as the response variable, one variable genotype of the focal plant, then one variable genotype of the “potmate” with 2 levels matching/mismatching, one variable herbivory with 2 levels = yes/no, then one variable pot, which you will use as a random factor. Then you can build the model which in R would look something like this: weight~herbivory * genotype * genotype of the “potmate”+(1|pot) Here you need to consider pot as a random factor.

  • $\begingroup$ One complexity that I can't wrap my head around: With a native-native pair, there is no "focal" plant. I'm not sure whether this matters for the model you propose or not (which otherwise makes sense to me). $\endgroup$
    – dipetkov
    Oct 16, 2023 at 13:15
  • $\begingroup$ Or do you mean that the unordered pair (Y_i1, Y_i2) in the i-th pot corresponds to two rows in the data table, one row where Y_i1 is focal and another row where Y_i2 is focal? $\endgroup$
    – dipetkov
    Oct 16, 2023 at 13:26
  • $\begingroup$ @dipetkov, I suppose it does not, since you would express the difference as an absolute value ? For the first proposed model, with the weight differences, there would be only one line per pot. I hope I got your question right ! $\endgroup$
    – CaroZ
    Oct 16, 2023 at 19:44
  • $\begingroup$ I think it's more interesting to model all the data. (The absolute value approach can't distinguish from the native genotype overcompeting the invasive one? Though admittedly that's contrary to the expectation.) So the second approach makes more sense to me, at least the mean part of the model. The variance part on the other hand, I'm not so sure. I've been wondering whether one effect of competition might not be to reduce the variance of the native genotype in the mixed-genotype pot. Anyway, +1 from me. $\endgroup$
    – dipetkov
    Oct 16, 2023 at 20:00
  • $\begingroup$ Then you can analyze whether variances are heterogenous or not. If they are, it would be better to use a glm, or a lm which takes heteroscedasticity into account. $\endgroup$
    – CaroZ
    Oct 16, 2023 at 20:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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