We conducted a trial evaluating efficiency of a lure and kill device called Magnet Med against Mediterranean fruit fly. We had three locations, in one location there were 2 orchards, one location with one orchard and in one location there were four orchards in which conducted the trial. We divided the property into two blocks one with Magnet Med and other without Magnet Med. Also some orchards had mix fruit trees such as pome and stone fruits some only with one type of fruits. In each treatment block we examined the fruits by selecting 30 trees randomly and from each tree 30 fruits were checked randomly for any strikes of fruit fly. In addition, we also setup several male and female traps and monitored fly numbers for several weeks at weekly interval.
I am not sure what type of analyses I should perform. I want to know if there is any difference in efficiency of Magnet med between orchards, between location, within and between crops.
Mr. Touhidur Rahman, I think you can easily analyse these data with the sofware SPSS. First devide the orchards as 1, 2, .....7 then compare with each other and also location can devide as the same way. But my opinion is, you should consider only the same fruit trees' data which are in the all locations or orchards. In case of within crops, I think you should need to devide into two categories as mixed and single crop. But for between crops, need to consider in same location not for all location togetherly as because so far I knew in some locations have only single crop. So this can not compare with mixed crops. The response of fruit fly from mixed crop might not be similar with single crop. Thanks
Thank you Mr Rahman. I appreciate your comments. I use either SAS, R or Genstat to analyse my data.
You have a very complex design with lots of factors. I suggest you should perform two separated linear mixed model with the factors: in A) location, orchards, mixed fuit, treatment, tree, damage and in B) location, orchards, mixed fuit, treatment, trap, time, amount of males, amout of females. In B you should set "time: trap" as a random effect. this will be your repeated measurement factor. If your data are not normaly distributed, you should perform a generalized linear mixed model, with the corresponting distribution. If you add the factors one by one in your model, you will see, if all factors are needed or just the treatment has an effect on your comparison between amout of flies or amout of damage in treated and untreated orchards. In R it is not that complicated to perform lmms. perhaps this will help: http://cran.r-project.org/web/packages/lme4/vignettes/Implementation.pdf, http://www.unil.ch/webdav/site/ee/shared/WorkshopD.pdf
good luck!
Thank you very much. Much appreciated. I have used REML in genstat with whole set of data. But I like your idea to use GLMM. I know how to perform GLMM in SAS but never used R. But the doc you provided seems very helpful to perform the analysis in R. I have R platform and will give a try. Thnak you again for your suggestions
Every analsis has pros and cons; analysis should not direct you, rather you direct analysis - in doing so, a quick presentation of data in pictorial form is a good way to go. When prresent your data on a graph, you may use a confidence interval for each of the factors that your are interested with. Hope it helps - you are most welcome to contact me personally.
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology