I recently ran a greenhouse study that had 3 different grass species growing across 4 different soil substrates. There were 40 of each soil:species combination, leading to 480 samples altogether.
We were curious in both germination and growth rates, so individuals were grown from seed in the substrates of interest, rather than first growing them in a separate medium. One of the species didn't germinate at all in 2 substrates, and the other species had generally low germination rates.
I measured leaf counts, and length of longest leaf 3 times per week, and took a dry aboveground biomass weight at the end.
It is my understanding that I cannot run a two-way ANOVA because of the strong discrepancies in my data (ie. dramatically different replicate counts)? What would be my best options for representing/analyzing this data?
Thank you
your experiment has 3 parts:
1. Main plot design is just 3X4 =12 plots .... anova...
2. each of these 12 combinations has a population of 40 (sort of sub-plot part). you need to model these 12 populations of size 40 e.g. take mean of each to get 12 means & analyse them in a 3X4 anova
3. you have plant growth & quality attributes observations. Again model these to generate summaries & other meaningfut functions. condense them down to 12 units of your design & do anova
you can't analyse the whole thing in one go...bring it all together when you write up...good luck..
Just a thought to check out with a statistician, instead of NA, list all dead plants as 0.0, meaning no growth or weight. Zeros can be part of an analysis, but NA can't be analyzed with numerical data.
Non germination is actually a result since that is what you were interested in, so you can analyse that. You might need to repeat the experiment using viable seeds since your germination rates may have been affected by low viability rather than the growing conditions
Andrew Paul McKenzie Pegman
The 0's being valuable for germination success makes sense. But doesn't that leave me short-changed for analysis of weights, because I'll have a few null values where there was never anything to measure?
I don't believe the missing combinations (ie. P soil + CC species, and CPS soil + CC species) can be weighed as 0g, because they never existed.
Yes. That’s why you should repeat the experiment with 100% viable seed since viability could be a confounding factor :)
- Badges
- Science topic
- Similar topics
- Agriculture
- Soil Science
- Soil