The problem of analyzing georeferenced cotton pest insect samples when a large percentage of the counts are zero is examined. The use of appropriate statistical methods for their analysis is required. To demonstrate this, georeferenced samples (n = 63) of tarnished plant bugs (TPBs; Lygus lineolaris [Palisot de Beauvois] (Heteroptera: Miridae)) were analyzed by three statistical methods and the results were compared. Correlation analysis of the sample counts with 25 classes of cotton growth derived from an unsupervized classification of multispectral imagery was followed by a complete enumeration analysis comprising three scenarios. The first scenario assumed the insect samples were unstratified. A distribution of sample averages was created by complete enumeration of all combinations of samples taken four at a time. The second scenario used imagery of the cotton fields to allocate the samples among three cotton growth categories (marginal, good or best) derived by a supervized classification of the 25 unsupervized classes. The insect samples associated with these categorical habitats were completely enumerated using allocations of 4, 6, 8 or 10 samples at a time from various sample sizes to determine how different allocations affected the results. The mean was not affected, but the standard deviation decreased with increased allocation sizes in all habitats. The third scenario used the two observers and three habitat categories to create six additional complete enumeration distributions by allocating four samples at a time from groups of varying sample sizes. These enumeration distributions are non-parametric estimators of the sampling distribution of: (1) the sample averages of a given sample size when samples are taken from the entire field, (2) the sample averages of a given size when samples are taken from each cotton habitat or (3) the sample averages of a given sample size from samples taken from each habitat by each observer. To support the enumeration analyses, these insect samples were analyzed further by Poisson regression models. These models showed significant differences between TPB counts by the two observers and among the habitats, whereas the observer by habitat interaction was not significant. For every combination of observer and cotton growth category, a Poisson regression model estimated the mean rate of TPB numbers. These means were similar to the corresponding modes of the complete enumeration distributions. The two non-standard methods showed that TPB numbers differed by habitat categories even though there were samples with a zero count, whereas a correlation analysis failed to identify a relationship between TPB sample counts and unsupervized habitat classes.
Precision Agriculture – Springer Journals
Published: Oct 10, 2008
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
All the latest content is available, no embargo periods.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud