Analysing Quantitative Genetic Data - Applied Statistics

course application:

Master students: please register using LSF

PhD students & Bachelor students: please send me an eMail with your student ID and your priority and I will ad you to the applicants' list: axel[dot]strauss[at]bio.lmu.de

Please note: due to the large number of requests we have changed the organisation a bit. We will work the first three days in a lab (D00.021) using course laptops. The second and third week we will move to G00.037.

We meet on Tuesday October 29th at 10:00am in D00.021.

Here is the time schedule.

NEW: This is my final list of book recommendations. There are surely more books and many of them are very good but these are the ones I have had a look in and they fit the common needs of my course participants.

Requirements:

• You need to have a Campus login to be able to use the desktop computers in the computer room as well as the course laptops.
• If you bring your own laptop, you need to (1) have access to Campus WiFi, (2) need to be able to login at genetics web page (Campus-Kennung), (3) need to have R installed (download here) and (4) need to have Excel or another spreadsheet application.

The course "Analysing quantitative genetic data - applied statistics" will provide its participants the statistical tools they need in daily lab life.

We will - step by step - find a way into statistical analysis of different kinds of data that you will probably deal with if working in genetics or during field work related to genetics based questions (basically all kind of data we collect here in the genetics department). There will also be enough time to explore R, what we will use for most analysis. No prior knowledge on statistics or R is needed (but of course helpful). If you contact me early enough (some days before the course) we can analyse your data during the course.

Here are some key points of the course. A time table will be available online before the course starts.

characteristic measures

We permanently summarise our data to get an overview of the results. Surely, the mean and standard deviation or standard error are the favourits. But are these actually good "representatives" of my data? Which other characteristic measures (Maßzahlen) are available, what do they mean, and which one to use best? (For example, median, quantiles, confidence interval,....). You will see how much the use of these measures depends on how your data look like. So we will also deal with the question "Should I use standard deviation or error on my bar plots?" (and we will extend it: "Should I use bar plots at all?").

data exploration

Based on what we have seen dealing with characteristic measures, we do some data exploration both in Excel and in R. You will get an introduction to R on this day so you will be trained before you need to apply it.

hypothesis testing

How do I ask scientific questions? What are null hypothesis, alternative hypothesis, and what does this "magic" p-value actually mean?

comparing two samples (data sets) I.

We will start statistics comparing two samples. You will get an introduction to t-test and Mann-Whitney-U-test and, based on the knowledge gained before, will apply it to both artificial data and real lab data.
We will also focus on some pitfalls, such as pseudoreplication (aka "technical replicates" in lab jargon), missing controls, or strange data distributions (e.g., outliers).

comparing more than two samples I.

This works just as the topic above but dealing with more complex data (more than two factors, factor with more than two levels).
We will first focus on ANOVA and Kruskal-Wallis-test and later on (simple) linear regression and the mix of both: ANCOVA. Also model evaluation (Have I been allowed to do ANOVA with my data?) will be part of this session.
Also important: what to report in thesis or publication?

data visualisation

Measurements and "p-values" etc. are important to present your results. However, graphics are what you need to visualise your data. There are some pitfalls and we will mostly deal with box-plots, strip-charts, scatter-plots, mosaic plots... (some of them later in the course when dealing with the respective type of data). We will do that mostly in R - step by step.

data transformation

If your data are of certain type (skewed distribution, heteroscedasticity, outliers, ...) that restricts the use of your preferred analysis, data transformation can make things possible. R provides very nice, easy tools to do that.

comparing more than two samples II.

Data transformation or the use of non-parametric analysis may not be needed if performing statistical analysis designed to deal with "strange data". So called "generalised linear models" can can be methods of choice if dealing with yes/no data, proportion data etc.

comparing two samples II.

If you wonder whether your observations follow a specific pattern/distributions (e.g., Mendel's law), Chi²-goodness of fit test (Chi²-Anpassungstest) will tell it to you. And if the question is whether observation patterns of character A are associated with those of character B, Chi² on contingency tables (Chi²-Homogenitätstest) will answer it.

data handling in R

While doing the different tests, you will learn how to access data in R. There are many nice functions making life easier that clicking hundreds of times in Excel (moving, subsetting, transforming, summarising data etc).

experimental design

During the whole course we will face issues of experimental design. Additionally, I will introduce power tests to you. These test can be helpful to see whether your sample size is sufficient.

multivariate analysis

We will have a short excursus to multivariate Analysis (Principal Component Analysis, Redundancy Analysis,...). You will most likely not need them in your lab life but it may be nice to know that these methods exist. We will not spent too much time with it.

We will switch permanently between lectures, "guided practical", exercises, and "free play". There will be enough time to play with data so you can also learn by "trial and error". We will work through everything by example and the use of real lab life data may help to understand and remember statistical possibilities.

It is not the main aim of the course to apply fancy high class statistics. If you are interested in special topics such as population genetics, have a look at the courses provided at the Institutes of Evolutionary Biology or Zoology.

During the course some sample data will be deposited here. Handouts will be available here as well.

Here are handouts and data 01 (I will bring prints of the working sheets)

Here are handouts and data 02 (I will bring prints of the working sheets)

Here are handouts and data 03 (I will bring prints of the working sheets) ^{– link removed due to copyright issues after removing login}

Here are handouts and data 04 (I will bring prints of the working sheets)

Here are handouts and data 05 (data and exercise)

Here is the last presentation