Introduction to Statistical Modelling

Mary Woodcock Kroble
Saturday 8 September 2012
Start date: 22 January 2013 - End date: 25 January 2013
Time: 9:00 am - 5:00 pm

Workshop full, waiting list only

The Centre for Research into Ecological and Environmental Modelling (CREEM) will be running a 4 day workshop to introduce participants to basic statistical modelling techniques in our purpose-built facilities at the University of St Andrews, Scotland. The course will be heavily practical based and in-class feedback will be provided using hand held personal response units (a.k.a clickers) to aid the learning of the workshop material.

Course content

To illustrate the methods, a marine mammal example will be used throughout the workshop and impact assessment examples will be used for the practicals. The software package R will be used in both taught and practical sessions.

The basics:Exploratory Data Analysis (EDA): Histograms Boxplots, Scatterplots,Measures of centre & spread: the mean, median & varianceConfidence intervals (CIs) & Hypothesis testing:o the t-test
o Analysis of Variance (ANOVA) & Adjusted confidence intervals for multiple comparisons

Linear models:Exploratory Data Analysis (EDA)Model Specification: Continuous terms & Factor variablesModel Fitting & Model SelectionModel Assessment:o The R-squared & Adjusted R-squared,
o Standardized residuals, Partial residual plots
o Formal tests for constant error variance, independence & normality
o Common fixes/alternativesCollinearity: Identifying collinearity (VIFs) and possible remediesModel Interpretation: parameter estimates, CIs for parameters, p-valuesPrediction: Confidence intervals & Prediction intervals

Extending the linear model: Generalized Least Squares:
Incorporating autocorrelation: empirical autocorrelation functions, autoregressive processes, choosing between structuresIncorporating non-constant error variance: modelling the funnel/fan effect

Models for proportions/binary outcomes: Generalized Linear Models (GLMs)
Estimating proportions
Confidence intervals for proportions: for large and small samplesChi-square tests for differences between proportionsGLMs for proportions: EDA, link functions, Fitting, Selection, Assessment, Parameter Interpretation

Models for count data: Generalized Linear Models (GLMs)
EDA, Confidence Intervals, Hypothesis TestsModel Specification (with and without an ‘offset’)
Model Selection & Assessment
Model Interpretation & Prediction

The R software package will be used for workshop-based practicals but all code will be provided and no prior experience with the R package is assumed.


For further information, please contact:

Centre for Research into Environmental and Ecological Modelling
University of St Andrews
The Observatory, Buchanan Gardens
St. Andrews
Scotland KY16 9LZ
Tel:+44 1334 461842
Fax: +44 1334 461800
Email: [email protected]

Share this story