New search
PNS0211
Advanced Statistics in Practice
Credits: 5 ECTS/HEC (each unit equals one credit; students can choose to participate in one or more units)
Course structure We will use a set of activities including initial readings, classroom discussions, hands-on activities, followed by home exercises. Each unit corresponds to one theme/workshop and is planned for one full day (8:45 - 17:00; including 1 hour lunch and 2 fika breaks, note exception Meta-analysis=2 days), equaling one credit (ca 27 hours of work for the PhD student). Example datasets will be provided, but students could use their own datasets as well.
Unit 1: Time series analysis. Teachers: Jonas Knape and Örjan Östman.
1 February 2021
1. Introduction to time series
What a time series is, how it differs from other data types, autocorrelation.
Basic time series models such as white noise, random walks and autoregressive models.
Connections to simple models of population dynamics, such as stochastic exponential growth.
2. AR, ARMA, structural changes, MAR, and multivariate (communities) statistics in time-series analysis.
3. State space models
4. Tools for time series analysis
5. Time series analysis of Swedish bird survey data.
Unit 2: GIS and spatial analysis. Teachers: Alistair Auffret and Mohammad Bahram. 15 February 2021.
1. Intro: Why GIS in R? Which packages?
2. Vector analysis: Buffer zones, Intersections, Summary statistics etc.
3. Raster analysis: Overlays, Reclassification, Raster calculator
4. Spatial autocorrelation and its effect on your results, strategies and methods for accounting or using spatial autocorrelation
5. Analysing spatially-structured ecological data
Unit 3: Meta-analysis Teacher: Julia Koricheva 12-13 April 2021
1. Formulating the research question for a meta-analysis
2. Data extraction from primary studies
3. Calculation of effect sizes
4. Combining effect sizes across studies using fixed and random effect models
5. Exploring causes of heterogeneity across studies (meta-regression)
6. Sensitivity analysis and testing for publication bias
Unit 4: Getting more out of community data Teachers: Tomas Roslin and Giovanni Strona. 3 May 2021
Matching the focus of the course on interpretation rather than methods, this unit will depart not from a specific method but from a series of ecological questions. Thus, we will start by defining fundamental questions on communities, then point to methods for approaching them:
1. Fundamental question: What can co-occurrence data reveal about ecological interactions?
Methodological application: How do we quantify co-occurrence? How do we separate a "real" pattern from one caused by chance alone (null model analysis)?
3(3)
2. Fundamental question: More generally: if there is structure, then what fundamental forces are behind it? What signals of such forces are hidden in the data?
Methodological application: How can the signals of community assembly be translated into a parameterized statistical model? (HMSC)
3. Fundamental question: Moving from patterns in the abundance and distribution of species to direct observations of who interacts with whom: can we identify the forces structuring interaction networks?
Methodological application: Analyses of trait-matching using recent techniques
Unit 5: Dealing with some complexities of GLMs Teacher: Matt Low. 31 May 2021
1. Understanding and comparing the Normal, Poisson and Binomial distributions.
2. What is overdispersion and what effect does it have on your results? How can it be quantified, corrected for, and what does it all mean biologically?
3. What is zero inflation? How zero-inflation can be thought of as a form of overdispersion (and hence dealt with using the previous solutions). How zero-inflation can be considered as a combination of two separate processes and modelled using either hurdle models or mixture models.
Course structure We will use a set of activities including initial readings, classroom discussions, hands-on activities, followed by home exercises. Each unit corresponds to one theme/workshop and is planned for one full day (8:45 - 17:00; including 1 hour lunch and 2 fika breaks, note exception Meta-analysis=2 days), equaling one credit (ca 27 hours of work for the PhD student). Example datasets will be provided, but students could use their own datasets as well.
Unit 1: Time series analysis. Teachers: Jonas Knape and Örjan Östman.
1 February 2021
1. Introduction to time series
What a time series is, how it differs from other data types, autocorrelation.
Basic time series models such as white noise, random walks and autoregressive models.
Connections to simple models of population dynamics, such as stochastic exponential growth.
2. AR, ARMA, structural changes, MAR, and multivariate (communities) statistics in time-series analysis.
3. State space models
4. Tools for time series analysis
5. Time series analysis of Swedish bird survey data.
Unit 2: GIS and spatial analysis. Teachers: Alistair Auffret and Mohammad Bahram. 15 February 2021.
1. Intro: Why GIS in R? Which packages?
2. Vector analysis: Buffer zones, Intersections, Summary statistics etc.
3. Raster analysis: Overlays, Reclassification, Raster calculator
4. Spatial autocorrelation and its effect on your results, strategies and methods for accounting or using spatial autocorrelation
5. Analysing spatially-structured ecological data
Unit 3: Meta-analysis Teacher: Julia Koricheva 12-13 April 2021
1. Formulating the research question for a meta-analysis
2. Data extraction from primary studies
3. Calculation of effect sizes
4. Combining effect sizes across studies using fixed and random effect models
5. Exploring causes of heterogeneity across studies (meta-regression)
6. Sensitivity analysis and testing for publication bias
Unit 4: Getting more out of community data Teachers: Tomas Roslin and Giovanni Strona. 3 May 2021
Matching the focus of the course on interpretation rather than methods, this unit will depart not from a specific method but from a series of ecological questions. Thus, we will start by defining fundamental questions on communities, then point to methods for approaching them:
1. Fundamental question: What can co-occurrence data reveal about ecological interactions?
Methodological application: How do we quantify co-occurrence? How do we separate a "real" pattern from one caused by chance alone (null model analysis)?
3(3)
2. Fundamental question: More generally: if there is structure, then what fundamental forces are behind it? What signals of such forces are hidden in the data?
Methodological application: How can the signals of community assembly be translated into a parameterized statistical model? (HMSC)
3. Fundamental question: Moving from patterns in the abundance and distribution of species to direct observations of who interacts with whom: can we identify the forces structuring interaction networks?
Methodological application: Analyses of trait-matching using recent techniques
Unit 5: Dealing with some complexities of GLMs Teacher: Matt Low. 31 May 2021
1. Understanding and comparing the Normal, Poisson and Binomial distributions.
2. What is overdispersion and what effect does it have on your results? How can it be quantified, corrected for, and what does it all mean biologically?
3. What is zero inflation? How zero-inflation can be thought of as a form of overdispersion (and hence dealt with using the previous solutions). How zero-inflation can be considered as a combination of two separate processes and modelled using either hurdle models or mixture models.
Syllabus and other information
Syllabus
PNS0211 Advanced Statistics in Practice, 5.0 Credits
Subjects
Biology Mathematic Statistics,Education cycle
Postgraduate levelGrading scale
Pass / Failed
Prior knowledge
Admitted to PhD-studies. Basic knowledge of statistics and RObjectives
This course aims to fill this gap by preparing students to analyze, interpret and report their data using the most up-to-date methods in R. Special attention will be given to discussion of questions springing from the students’ own work and the biological interpretation of data. The course is intended to deepen the students’ understanding of all aspects of ecological inferences, not as a cookbook of "how to". Learning outcomes By the end of the course, students are expected to be able to 1- Demonstrate the ability to identify relevant functions and packages in R for analyzing their own datasets. 2- Analyze data using R, including reporting advanced statistics, univariate and multivariate modelling and generation of graphs. 3- Interpret, think critically and draw conclusions on data analysis results.Content
Credits: 5 ECTS/HEC (each unit equals one credit; students can choose to participate in one or more units) Course structure We will use a set of activities including initial readings, classroom discussions, hands-on activities, followed by home exercises. Each unit corresponds to one theme/workshop and is planned for one full day (8:45 - 17:00; including 1 hour lunch and 2 fika breaks, note exception Meta-analysis=2 days), equaling one credit (ca 27 hours of work for the PhD student). Example datasets will be provided, but students could use their own datasets as well. Unit 1: Time series analysis. Teachers: Jonas Knape and Örjan Östman. 1 February 2021 1. Introduction to time series What a time series is, how it differs from other data types, autocorrelation. Basic time series models such as white noise, random walks and autoregressive models. Connections to simple models of population dynamics, such as stochastic exponential growth. 2. AR, ARMA, structural changes, MAR, and multivariate (communities) statistics in time-series analysis. 3. State space models 4. Tools for time series analysis 5. Time series analysis of Swedish bird survey data. Unit 2: GIS and spatial analysis. Teachers: Alistair Auffret and Mohammad Bahram. 15 February 2021. 1. Intro: Why GIS in R? Which packages? 2. Vector analysis: Buffer zones, Intersections, Summary statistics etc. 3. Raster analysis: Overlays, Reclassification, Raster calculator 4. Spatial autocorrelation and its effect on your results, strategies and methods for accounting or using spatial autocorrelation 5. Analysing spatially-structured ecological data Unit 3: Meta-analysis Teacher: Julia Koricheva 12-13 April 2021 1. Formulating the research question for a meta-analysis 2. Data extraction from primary studies 3. Calculation of effect sizes 4. Combining effect sizes across studies using fixed and random effect models 5. Exploring causes of heterogeneity across studies (meta-regression) 6. Sensitivity analysis and testing for publication bias Unit 4: Getting more out of community data Teachers: Tomas Roslin and Giovanni Strona. 3 May 2021 Matching the focus of the course on interpretation rather than methods, this unit will depart not from a specific method but from a series of ecological questions. Thus, we will start by defining fundamental questions on communities, then point to methods for approaching them: 1. Fundamental question: What can co-occurrence data reveal about ecological interactions? Methodological application: How do we quantify co-occurrence? How do we separate a "real" pattern from one caused by chance alone (null model analysis)? 3(3) 2. Fundamental question: More generally: if there is structure, then what fundamental forces are behind it? What signals of such forces are hidden in the data? Methodological application: How can the signals of community assembly be translated into a parameterized statistical model? (HMSC) 3. Fundamental question: Moving from patterns in the abundance and distribution of species to direct observations of who interacts with whom: can we identify the forces structuring interaction networks? Methodological application: Analyses of trait-matching using recent techniques Unit 5: Dealing with some complexities of GLMs Teacher: Matt Low. 31 May 2021 1. Understanding and comparing the Normal, Poisson and Binomial distributions. 2. What is overdispersion and what effect does it have on your results? How can it be quantified, corrected for, and what does it all mean biologically? 3. What is zero inflation? How zero-inflation can be thought of as a form of overdispersion (and hence dealt with using the previous solutions). How zero-inflation can be considered as a combination of two separate processes and modelled using either hurdle models or mixture models.Additional information
Register for the course (all or separate units) by sending an email to the course organizer Mohammad Bahram (mohammad.bahram@slu.se) no later than November 29 and indicate what units you are interested in.Location: The course is planned to be delivered either in real life (given an end to social distancing) in Ultuna campus or online (Slack/Zoom) depending on the pandemic situation.