Program Codes

A diagnostic analysis for exploring genotype-by-environment interactions which often are found in plant breeding trials. Example programs given for a national rapeseed trial. Codes written for the SAS programming language.

• Create simulation data (sas)
• Generalized linear mixed model (LaPlace) (sas)
• Process likelihood model (sas)

The model is capable of working on large scales and can be adjusted for various factors. Command line routines written in ANSI C and the Perl scripting language.

Estimation of doses in an unknown sample assuming the calibration curve known or estimated. Codes written in SAS.

Dose-response models are common in agriculture and biology. Dose-response curves are often used in bioassay to determine unknown dosage levels. The programs below illustrate the process of unknown dose estimation under two conditions:

1. Calibration curve is known without error
2. Calibration curve is estimated with known error

Examples are given for the logistic model:

y = 1/(1 + exp(-B*(dose - G)))

where y = the proportion of successes, B is a rate related parameter, G is the estimate of the 50th percentile and dose is the applied dose.

The Bayesian solution reparametrized the first form in terms of T1, the initial proportion of successes at dose = 0, and T2, the final proportion of successes at dose = maximum. Prior distributions for T1 and T2 are assumed uniform.

Source

• Calibration curve known (sas)
• Calibration curve estimated (sas)

Black Vine Weevil data (txt)

Command line routines written in ANSI C and the Perl scripting language.

• Bayes differences (zip)
• Bayes error matrix (zip)

These programs compute Binomial, Bayesian and Bootstrap estimations of the omissional and commission error rates used in assessing the reliability of remotely sensed imagery. Command line routines written in ANSI C and the Perl scripting language.

Examples are given for Common Crupina, a weed found in the northwestern United States. Codes written for the SAS programming language.

Various estimation techniques: Probit, NLIN, GNLIM and Bayesian. Example programs given in SAS for egg hatch in black vine weevil. Codes written in SAS.

Dose-response models are common in agriculture and biology. The programs demonstrate various techniques of binary dose-response estimation within SAS. Examples are given for the logistic mModel:

y = 1/(1 + exp(-B*(dose - G)))

where y = the proportion of successes, B is a rate related parameter, G is the estimate of the 50th percentile and dose is the applied dose. The probit model analyzes a linearized form:

log(y/(1-y)) = A` - B*dose

Where A` is B*G from the previous form. The Bayesian solution reparametrized the first form in terms of T1, the initial proportion of successes at dose = 0, and T2, the final proportion of successes at dose = maximum. prior distributions for T1 and T2 are assumed uniform.

Source

• Probit analysis (sas)
• NLIN analysis (sas)
• GNLIM analysis (sas)
• Bayesian analysis (sas)
• GNLIM treatment comparison analysis (sas)

Continuous Data

• Example 1 (sas) | Wheat ALS data (xls)

Poisson/Negative Binomial Count Data

• Example 2 (sas) | Dry pea data (xls) uses variety “C” only

Poisson/Negative Binomial Count Data — Variety Comparison

• Example 3 (sas) | Dry pea data (xls)* (*notes same file listed above)

Binomial Data

• Example 4 (sas) | Lettuce germination data (xls)

PowerPoint presentations by Seefeldt, Shafii and Price

• Dose-response: background and perspectives on the development of analysis methodology (ppt)
• Estimation techniques for dose-response functions (ppt)
• Applied of dose-response models in weed science (ppt)

Code (SAS) and data (Excel) for examples (*notes same file listed above)

• Probit Analysis (sas) | Vernalization data (xls)
• NLIN analysis (sas) | Vernalization data (xls)*
• NLMIXED analysis (sas) | Vernalization data (xls)*
• NLMIXED analysis-Seefeldt (sas) | Wild oat data (xls)
• Fitting poisson data-peas (sas) | Harmony pea data (xls)
• Fitting exponential model-peas (sas) | Harmony pea data (xls)*
• Model reparameterization-Seefeldt (sas) | Wild oat data (xls)*
• Comparison of treatments-Seefeldt (sas) | Wild oat data (xls)*
• Fitting random effects-gnat (sas) | Fungus gnat data (xls)
• Bayesian estimation-lettuce (sas) | Lettuce data (xls)
• Bayesian nonparametric bioassay estimation (pdf)

Command line routines written in ANSI C and the Perl scripting language.

Simple exponential model (sas)

Segmented exponential model (sas)

Weibull model

• Onion seed data (xls)
• Weibull model (sas)
• Profile t-plot (sas)
• Profile pair sketches (sas)
• Treatment comparison — nlin (sas)
• Treatment comparison — nlmixed (sas)

Modified exponential model (sas)

Bayesian log-logistic model (sas)

Random effects model (sas)

Useful references

• Bates, D.M., and D.G. Watts. 1988. Nonlinear Regression Analysis and its Applications. John Wliey and Sons. New York.
• Finney, D. J. 1971. Probit Analysis. Cambridge University Press, London.
• Ratkowsky, D. A. 1989. Handbook of Nonlinear Regression Models. Marcel Dekker, Inc. 241 pp.
• SAS Inst. Inc. 2004. SAS OnlineDoc, Version 9, Cary, NC.

Requires no assumptions on sample sizes or parent distributions. Example programs given for a dairy cow study. Codes written in the ANSI C language under the GPL license.

Quadratic discriminant function which uses prior information for yellow starthistle from spatial landscape information. Provides a better classification of yellow starthistle than a uniform prior. Sample images and program given. Large images will require substantial computing resources. Codes written in the ANSI C language under the GPL license.