Hi there,
I'm designing an experiment (not necessarily to be conducted) in which the primary source of nitrogen within large monocultures is manipulated, and yield of crop is measured over time. The objective is to determine the feasibility of using peanuts as a primary source of nitrogen on large-scale agricultural operations.
Site: large-scale (~200 ha) orange grove in Florida.
Treatments: one plot (1-ha) with only peanuts, one plot (1-ha) with some peanuts plus strategically applied synthetic fertilizer.
Control: one plot (1-ha) with no change in fertilizer regime.
Data will be collected from 5 subplots (400m2) per plot (15 total).
Recorded from each subplot will be number of trees and number of boxes of oranges yielded per tree. This data will be scaled up to estimate number of trees per hectare and number of boxes of oranges yielded per hectare.
Feasibility would be determined by change in yield over time. Little variation in average monthly, seasonal, and overall yield per plot would indicate potential feasibility of using peanuts as fertilizer. Great variation between experimental plots and control plot may indicate that peanuts are not suitable sources of nitrogen for monoculture oranges.
What would be the appropriate statistical test to analyze this data? I'm thinking a single-factor ANOVA will tell me differences between treatments, but how can I compare data per treatment over time to determine overall change? Repeated measures ANOVA?
Any suggestions/examples would be appreciated.
Thank you.
I can't comment on the stats but would suggest that you need a N-balance; consequently, you need measures to derive that.
For applying the same, you need 12 degree of freedom. You have 3 treatments. 5 subplots can be treated as replication. Then your df is only 8. If you have 4 treatments, it would be good.
Expecting comments from the statisticians.
Interesting. You are interested in trends; the Jonckheree-Terpstra test, although it is a non-parametric test, can give you a trends test.
Hi Candice, the statistic depends on your research question. if you are interested in evaluating treatment differences over time, I believe the repeated measurement ANOVA is appropriate, with time as the repeated measure. You will need to see which covariance structure fits better in your model, which accounts for the fact that measurements made near in time are more correlated than measurements made further apart. For example, orange yield in March would be more correlated with yield in April than yield in September (I don't know much about orange production, this is just an example).
You also need to think about the practical interpretation of your results. For example, if you use month as repeated measure and your response variable is yield, your inference will be about how treatments affected yields over months (maybe different trends for each treatment?, when to reapply fertilizers based on declining yields?)
I hope this helped.
I think you should try an Analysis of variance for repeated measures (rmANOVA), including statistical significance for the effects of treatment (between-subjects factor) and sampling time (within-subjects factor), as well as the interactions between them. When the assumption of sphericity is violated, the degrees of freedom are corrected using Greenhouse-Geisser estimates of sphericity.
i suggest that you may have a correlation between effect of fertiliers to its yield.Also a series of test pertaining to its soilphysico chemical analaysis showing soil fertility when applied with your fertilizer. Further a multifactorial design will also be enployed based on your undertakings
- Badges
- Science topic
- Similar topics
- Geoscience
- Atmospheric Sciences
- Climate Science
- Climate Change