Case studies




A survey of cattle keeping practices of the Orma tribe in eastern Kenya and levels of milk production

Pilot household survey combining questionnaire and quantitative measurement approaches. Demonstrates analytical methods for fitting regression lines with different slopes. Reports methods for the presentation of results of regression analysis. Discusses suitability of survey approach. Describes additional study that uses participatory methods.


Identification of 'elite' Ankole cattle in a herd-monitoring study in Uganda

Describes design of observational study. Least squares hierarchical analyses of variance undertaken to estimate herd effects. Data then corrected for these effects and residuals sorted to identify top yielding cows. GenStat programming capabilities demonstrated for calculating total milk yield.


Exploratory analysis of factors affecting weaning weight of Dorper and Red Maasai lambs

Exploratory methods demonstrated for evaluating relationships with weaning weight for variables such as year of birth, sex, age at weaning and age of dam. Methods of subdividing the sums of squares to evaluate alternative models described. Alternative forms of presentation of results from general least squares analyses of variance discussed.


Mixed model analysis for the estimation of components of genetic variation in lamb weaning weight

Demonstrates use of REML for mixed model with random effects for sire and dam. Multi-level structure of the data explored. Method for expressing different layers diagrammatically in the form of a 'mixed model tree' shown. Output from REML described for split-plot and compared with output from method of general least squares. Interpretation of REML output discussed.


Research strategy for selecting 'best-bet' accessions of Napier grass for feeding to livestock

Research strategy and research process laid out for designed experiments and across site trials. Series of studies planned to meet objective described. Randomised block field and sheep experiments. Row-column designs discussed. Multi-location trial. Important ingredients of a study protocol illustrated. Principal component analysis. Cluster analysis.

6 Field studies to evaluate the benefits on milk production of providing smallholder farmers in Kenya access to credit to allow shifting of concentrate feeding to early lactation

Research strategy and research process described in relation to on-farm, cross-sectional and participatory longitudinal studies and concurrent farmer interviews. Problem of collapse of longitudinal study design and resulting complications in data analysis discussed. Data management issues discussed. Adequacy of statistical methods used in published results investigated. Methods of multiple regression and correlation analysis applied. Programming capabilities of GenStat for replicating calculations demonstrated.

7 Exploring daily rainfall data to investigate evidence of climate change in southern Zambia and its implication for farmers in the area

Exploratory and descriptive analysis for analysis of daily rainfall data presented with Instat. Illustrates problems in managing and editing climatic data. Illustrates importance of keeping data logs. Methods for reporting and disseminating results to farmers through plays and radio broadcasts described.

8 An analysis of the number of goats needed in a feeding experiment

Attempts to rectify experiment that went wrong. Randomised block (batch) experiment. Analysis of covariance with pre-treatment period as covariate. Sample size estimation.

9 Design and analysis of experiments on pollen grain germination and viability rates of the leguminous Sesbania sesban (L.) Merr tree species

Multi-level design to compare pollen germination rates across a range of sucrose concentrations. Analysis of variance with polynomial regression. Stacking of data. Estimation of variance components. Improvements in experimental design investigated. Distribution of sample in hierarchical studies discussed

10 Impact assessment of the effect of tsetse control on livestock productivity

Describes longitudinal intervention monitoring study with periods without tsetse control followed by periods with tsetse control. Discusses problems of interpretation of 'before-after' studies. Data management of longitudinal data discussed. Describes methods of data reduction to form 6-month observational units. Uses general least squares and REML methods to adjust for other factors. Videos included demonstrating data collection and research in field.

11 Analysis of cattle production system data from the Swaziland Farm Animal Genetic Resources Survey

Design of a questionnaire for a survey on livestock illustrated and design of survey described. Describes data management system. Demonstrates, using both Instat and GenStat, methods for handling data from surveys, in particular multiple response data. Data exploration with tables and graphs to determine patterns, trends and distributions. Confirmation of exploratory results where conclusions not clear by Pearson's chi-square test and REML. Method for ranking responses illustrated. Methods of reporting discussed.

12 Swaziland Farm Animal Genetic Resources Survey: estimation of population numbers for different livestock species

Estimates cattle population numbers and their standard errors from clustered sample survey of homesteads. Compares formulae for situations where population sizes for sampling units are known and unknown. Sampling strategies with equal sample sizes for each sampling unit and sizes proportional to unit size compared. Discusses importance of randomisation and effects of different types of sampling on reliability of population estimates.

13 Design of an experiment to compare performance and meat quality characteristics of three Ethiopian goat genotypes under varying nutritional conditions

Describes how to prepare study protocol. Discusses study design for factorial experiment.

14 An epidemiological study of factors affecting the prevalence of schistosomiasis (bilharzia) along the River Kochi in the West Nile region of Uganda

MSc epidemiological observational study. Lessons on design and management of data recording sheets. Approaches to exploratory analysis. Regression analysis. Logistic regression.

15 A factorial experiment to study the effect of seeding rate and nitrogen side-dressing on yields of two dry bean (Phaseolus vulgaris L.) cultivars

Randomised block 3x2x2 factorial experiment. Plot design and methods of sampling. Split-plot-in-time analysis of weekly values. Use of contrasts in analysis of variance. Recording sheet preparation. Residual plot analysis. Interpretation of standard errors in split plot analysis. Reporting results from factorial experiments.

16 Influence of planting date and organic fertilisation on growth, yield and corm quality of South African taro landraces

Research strategy and study design. Series of studies planned to meet objectives. Plot number arrangements in field experiments. 4x3x3 split-plot factorial experiment with 4 planting dates. 2x3x3x3 split-split plot experiment. Skeleton analyses of variance. Use of GenStat to design experimental layout. Exploratory analysis. Residual plot analysis. Understanding zero values - missing or not. Split-plot analysis of variance. Reporting results from factorial experiments.

17 The effect of spent yeast (Saccharomyces cerevisiae) strain SAB5 on the performance of village chickens

BSc study in poultry feeding. Experimental design error. Statistical analysis of group versus individual feeding experiments. Sample size considerations.