MSE FINDR: A Shiny R Application to Estimate Mean Square Error Using Treatment Means and Post Hoc Test Results.

Research synthesis methods such as meta-analysis rely primarily on appropriate summary statistics (i.e., means and variance) of a response of interest for implementation to draw general conclusions from a body of research. A commonly encountered problem arises when a measure of variability of a response across a study is not explicitly provided in the summary statistics of primary studies. Typically, these otherwise credible studies are omitted in research synthesis, leading to potential small-study effects and loss of statistical power. We present MSE FINDR, a user-friendly Shiny R application for estimating the mean square error (i.e., within-study residual variance, [Formula: see text]) for continuous outcomes from analysis of variance (ANOVA)-type studies, with specific experimental designs and treatment structures (Latin square, completely randomized, randomized complete block, two-way factorial, and split-plot designs). MSE FINDR accomplishes this by using commonly reported information on treatment means, significance level (α), number of replicates, and post hoc mean separation tests (Fisher's least significant difference [LSD], Tukey's honest significant difference [HSD], Bonferroni, Sidák, and Scheffé). Users upload a CSV file containing the relevant information reported in the study and specify the experimental design and post hoc test that was applied in the analysis of the underlying data. MSE FINDR then proceeds to recover [Formula: see text] based on user-provided study information. The recovered within-study variance can be downloaded and exported as a CSV file. Simulations of trials with a variable number of treatments and treatment effects showed that the MSE FINDR-recovered [Formula: see text] was an accurate predictor of the actual ANOVA [Formula: see text] for one-way experimental designs when summary statistics (i.e., means, variance, and post hoc results) were available for the single factor. Similarly, [Formula: see text] recovered by the application accurately predicted the actual [Formula: see text] for two-way experimental designs when summary statistics were available for both factors and the sub-plot factor in split-plot designs, irrespective of the post hoc mean separation test. The MSE FINDR Shiny application, documentation, and an accompanying tutorial are hosted at https://garnica.shinyapps.io/MSE_FindR/ and https://github.com/vcgarnica/MSE_FindR/. With this tool, researchers can now easily estimate the within-study variance absent in published reports that nonetheless provide appropriate summary statistics, thus enabling the inclusion of such studies that would have otherwise been excluded in meta-analyses involving estimates of effect sizes based on a continuous response.

Into the Trees: Random Forests for Predicting Fusarium Head Blight Epidemics of Wheat in the United States.

Constructing models that accurately predict Fusarium head blight (FHB) epidemics and are also amenable to large-scale deployment is a challenging task. In the United States, the emphasis has been on simple logistic regression (LR) models, which are easy to implement but may suffer from lower accuracies when compared with more complicated, harder-to-deploy (over large geographies) model frameworks such as functional or boosted regressions. This article examined the plausibility of random forests (RFs) for the binary prediction of FHB epidemics as a possible mediation between model simplicity and complexity without sacrificing accuracy. A minimalist set of predictors was also desirable rather than having the RF model use all 90 candidate variables as predictors. The input predictor set was filtered with the aid of three RF variable selection algorithms (Boruta, varSelRF, and VSURF), using resampling techniques to quantify the variability and stability of selected variable sets. Post-selection filtering produced 58 competitive RF models with no more than 14 predictors each. One variable representing temperature stability in the 20 days before anthesis was the most frequently selected predictor. This was a departure from the prominence of relative humidity-based variables previously reported in LR models for FHB. The RF models had overall superior predictive performance over the LR models and may be suitable candidates for use by the Fusarium Head Blight Prediction Center.

Accuracy in the prediction of disease epidemics when ensembling simple but highly correlated models.

Ensembling combines the predictions made by individual component base models with the goal of achieving a predictive accuracy that is better than that of any one of the constituent member models. Diversity among the base models in terms of predictions is a crucial criterion in ensembling. However, there are practical instances when the available base models produce highly correlated predictions, because they may have been developed within the same research group or may have been built from the same underlying algorithm. We investigated, via a case study on Fusarium head blight (FHB) on wheat in the U.S., whether ensembles of simple yet highly correlated models for predicting the risk of FHB epidemics, all generated from logistic regression, provided any benefit to predictive performance, despite relatively low levels of base model diversity. Three ensembling methods were explored: soft voting, weighted averaging of smaller subsets of the base models, and penalized regression as a stacking algorithm. Soft voting and weighted model averages were generally better at classification than the base models, though not universally so. The performances of stacked regressions were superior to those of the other two ensembling methods we analyzed in this study. Ensembling simple yet correlated models is computationally feasible and is therefore worth pursuing for models of epidemic risk.

Disease Risk, Spatial Patterns, and Incidence-Severity Relationships of Fusarium Head Blight in No-till Spring Wheat Following Maize or Soybean.

The first large-scale survey of Fusarium head blight (FHB) in commercial wheat fields in southern Brazil was conducted over three years (2009 to 2011). The objectives were to: (i) evaluate whether increased FHB risk is associated with within-field maize residue; (ii) determine the spatial pattern of FHB incidence; and (iii) quantify the relationship between FHB incidence and severity. FHB was assessed in a total of 160 fields between early milk and dough. Incidence ranged from 1.0 to 89.9% (median = 25%) and severity from 0.02 to 18.6% (median = 1.3%). FHB risk was neither lower nor higher in wheat following maize than in wheat following soybean. Only 18% of fields were classified as having aggregated patterns of FHB-symptomatic spikes. A binary power law description of the variances was consistent with an overall random pattern of the disease. These results conform with the hypothesis that FHB epidemics in southern Brazil are driven by sufficient atmospherically-transported inoculum from regional sources. The incidence-severity relationship was coherent across growing season, growth stage, and previous crop; one common fitted curve described the relationship across all observations. Estimating severity from incidence may be useful in reducing the workload in epidemiological surveys.

A machine learning interpretation of the contribution of foliar fungicides to soybean yield in the north-central United States.

Foliar fungicide usage in soybeans in the north-central United States increased steadily over the past two decades. An agronomically-interpretable machine learning framework was used to understand the importance of foliar fungicides relative to other factors associated with realized soybean yields, as reported by growers surveyed from 2014 to 2016. A database of 2738 spatially referenced fields (of which 30% had been sprayed with foliar fungicides) was fit to a random forest model explaining soybean yield. Latitude (a proxy for unmeasured agronomic factors) and sowing date were the two most important factors associated with yield. Foliar fungicides ranked 7th out of 20 factors in terms of relative importance. Pairwise interactions between latitude, sowing date and foliar fungicide use indicated more yield benefit to using foliar fungicides in late-planted fields and in lower latitudes. There was a greater yield response to foliar fungicides in higher-yield environments, but less than a 100 kg/ha yield penalty for not using foliar fungicides in such environments. Except in a few production environments, yield gains due to foliar fungicides sufficiently offset the associated costs of the intervention when soybean prices are near-to-above average but do not negate the importance of disease scouting and fungicide resistance management.

Yield Loss in Sweet Corn Caused by Puccinia sorghi: A Meta-Analysis.

Data sets meeting established criteria were included in a meta-analysis of the relationship between percent common rust severity and percent relative yield loss in sweet corn (processing: 20 data sets; fresh market: 14 data sets). The slope of the linear, zero intercept relationship was estimated from each data set. Overall slopes and their respective 95% confidence intervals for the processing and fresh market situations were estimated by a random effects meta-analysis. Results indicated that for processing sweet corn, every 10% increase in rust severity reduced yield by 2.4 to 7.0%; the corresponding reduction for fresh market sweet corn was between 3.0 and 6.2%. A meta-regression analysis did not identify any factors that could account for the observed variability between data sets. An expression was then obtained for Δs, the reduction in rust severity a single strobilurin fungicide spray ought to cause for the cost of the treatment to be offset by the value of the resulting yield improvement. The empirical distribution of Δs,was derived by stochastic simulation, which showed that fungicide usage could be cost effective 90% of the time when rust severity is reduced by 12% in processing sweet corn and by 5% in fresh market sweet corn.

Incidence, Spatial Patterns, and Associations Among Viruses in Snap Bean and Alfalfa in New York.

Recent epidemics in snap bean (Phaseolus vulgaris) characterized by virus-like symptoms prompted a survey of commercial fields for Alfalfa mosaic virus (AMV), Cucumber mosaic virus (CMV), and the Bean yellow mosaic virus (BYMV)/Clover yellow vein virus (ClYVV) complex in 2002 and 2003. Snap bean fields were either remote from or adjacent to alfalfa (Medicago sativa), a putative source of these viruses. Bean fields were sampled at the bloom stage in both years. Model-adjusted mean incidences of infection by AMV, BYMV/ClYVV, and CMV were 41.96, 6.56, and 6.69%, respectively, in alfalfa, and 6.66, 6.38, and 17.20% in snap bean. In 2002, 25.9% of snap bean plants were infected by more than one virus; <1% had more than one virus in 2003. Virus incidences did not differ between snap bean adjacent to or remote from alfalfa, but incidence of infection by AMV and BYMV/ClYVV was significantly higher in snap bean planted later in the season rather than earlier. In 2002, there was a positive association between AMV and CMV in the tendency to find both viruses in the same snap bean plant. In some years, infection by aphid-transmitted viruses can become widespread in snap bean in New York.

Spatial Patterns of Viable Spore Deposition of Gibberella zeae in Wheat Fields.

ABSTRACT An increased understanding of the epidemiology of Gibberella zeae will contribute to a rational and informed approach to the management of Fusarium head blight (FHB). An integral phase of the FHB cycle is the deposition of airborne spores, yet there is no information available on the spatial pattern of spore deposition of G. zeae above wheat canopies. We examined spatial patterns of viable spore deposition of G. zeae over rotational (lacking cereal debris) wheat fields in New York in 2002 and 2004. Viable, airborne spores (ascospores and macroconidia) of G. zeae were collected above wheat spikes on petri plates containing a selective medium and the resulting colonies were counted. Spores of G. zeae were collected over a total of 68 field environments (three wheat fields during 54 day and night sample periods over 2 years) from spike emergence to kernel milk stages of local wheat. Spatial patterns of spore deposition were visualized by contour plots of spore counts over entire fields. The spatial pattern of spore deposition was unique for each field environment during each day and night sample period. Spore deposition patterns during individual sample periods were classified by spatial analysis by distance indices (SADIE) statistics and Mantel tests. Both analyses indicated that the majority (93%) of the spore deposition events were random, with the remainder being aggregated. All of the aggregated patterns were observed during the night. Observed patterns of spore deposition were independent of the mean number of viable spores deposited during individual sample periods. The spatial pattern for cumulative spore deposition during anthesis in both years became aggregated over time. Contour maps of daily and cumulative spore deposition could be compared with contour maps of FHB incidence to gain insights into inoculum thresholds and the timing of effective inoculum for infection.

Sampling for the incidence of aphid-transmitted viruses in snap bean.

ABSTRACT Data collected in 2002 and 2003 on Alfalfa mosaic virus and Cucumber mosaic virus incidences of infection in commercial snap bean fields in New York State were used to develop relationships between disease incidence (p(low)) and sample size while accounting for the inherent spatial aggregation of infected plants observed with these two viruses. For a plan consisting of 300 sampled plants (N = 60 quadrats, n = 5 plants per quadrat), estimating p(low) from the incidence of positive groups (p(high); testing of N = 60 grouped samples) provides the same precision in p(low) as testing 200 plants individually, up to about p(low) = 0.2. Above that, the confidence interval width for p(low) obtained via group testing becomes markedly larger than the width obtained by testing individual plants. Our results suggest using group testing until p(high) is in the range [0.35, 0.59], which corresponds to p(low) in [0.1, 0.2]. Results indicate that group testing can be more economical than the testing of individual plants without loss of precision, at lower incidences of infection. The approach presented provides a general framework for sampling and the estimation of incidence of other aphid-transmitted viruses in snap bean.

A Rainfall-Based Model for Predicting the Regional Incidence of Wheat Seed Infection by Stagonospora nodorum in New York.

ABSTRACT Our goal was to develop a simple model for predicting the incidence of wheat seed infection by Stagonospora nodorum across western and central New York in any given year. The distribution of the incidence of seed infection by S. nodorum across the region was well described by the beta-binomial probability distribution (parameters p and theta). Mean monthly rainfalls in May and in June across western and central New York were used to predict p. The binary power law was used to predict theta. The model was validated with independent data collected from New York. The predicted distribution of seed infection incidence was not statistically different from the actual distribution of the incidence of seed infection.

Differential Seed Infection of Wheat Cultivars by Stagonospora nodorum.

Seed of soft white winter wheat collected from New York regional cultivar trials in 1995 and 1996 were assayed on an agar medium selective for Stagonospora nodorum. Incidence of seed infection varied with production environment. Relative incidence of seed infection differed significantly among cultivars and was consistent across environments. The flag leaves and ears of 12 cultivars were inoculated quantitatively at flowering in a glasshouse. Cultivars did not differ significantly in disease on the flag leaves. Incidence of seed infection for all cultivars was above 60%, but was significantly lower in Delaware and Houser than in other cultivars. Results confirm that wheat cultivars differ in their relative susceptibility to seed infection by S. nodorum. Resistance in wheat to seed infection by S. nodorum may be a useful mechanism for reducing initial inoculum in areas where infected seed is considered the primary inoculum source for Stagonospora nodorum blotch.

Modeling temporal trends in aphid vector dispersal and cucumber mosaic virus epidemics in snap bean.

Cucumber mosaic virus (CMV) has become a major limiting factor in snap bean production in the Great Lakes region of North America, and epidemics have occurred more frequently since the soybean aphid, Aphis glycines Matsumura, was introduced. Major aphid vectors of CMV epidemics were identified by statistically relating their temporal dispersal trends to the incidence of CMV. Alates were monitored weekly using water pan traps in 74 snap bean fields in New York and Pennsylvania from 2002 to 2006. Plants were tested for CMV by ELISA one time during late bloom in 2002 and 2003 and weekly over the season from 2004 to 2006. Principal vectors of CMV included Acyrthosiphon pisum (Harris), A. glycines, Aphis gossypii Glover, and Therioaphis trifolii (Monell). Among these, A. glycines and T. trifolii were likely responsible for severe CMV epidemics because they were among the most abundant species captured, they efficiently transmit CMV, and their dispersal activity was positively correlated with periods when CMV incidence was highest. Moreover, because high numbers of A. glycines and T. trifolii disperse during July and August, snap bean fields planted beyond late June are at risk for infection during early vegetative stages and are subsequently more at risk for yield loss. In contrast, plantings up to late June are less likely to become infected during early developmental stages and should escape yield loss because major vectors are dispersing infrequently. CMV-resistant or tolerant snap bean varieties should be planted after late June to reduce the risk of yield loss.