Statistics colloquium - general information

We are offering a statistics colloquium for linguists conducting experimental research to provide a platform to analyze and discuss experimental data together. Meetings are on thursdays, 2 p.m., on a semi-regular basis, depending on demand. If you are interested in participating, please write an email to emilia dot ellsiepen at linguist dot jussieu dot fr. This colloquium addresses primarily former participants of the statistics course and linguistis with a background in inferential statistics. Click here for updates and materials of individual sessions.

General goals

• provide individual assistance with the data analysis for the experimenting linguist
• give all participants the chance to get hands-on experience with different data types and experimental designs

How to participate

1. Contributing your own data

• notify Emilia (emilia dot ellsiepen at linguist dot jussieu dot fr) that you would like to contribute your experimental data (and when)
• provide general information about the experiment
• send in the data itself (please see data format below)

2. Participating

• take part in the doodle poll and/or signal your interest directly by email (emilia dot ellsiepen at linguist dot jussieu dot fr)
• if possible, always bring your lap-top to be able to run analyses yourself

Data format

Since reformatting the data to meet the requirements (by r, but also more generally) often takes up a lot of time, we strongly encourage to format the data beforehand in the following way: (in r, you can use the package reshape2 for this)

One data point per line: a data point usually corresponds to an individual measurement. So for example for an acceptability judgment experiment, each line should contain exactly one rating. In the case of multiple measurements (e.g. in a self-paced reading experiment with more than one region of interest), the different measurements should occupy different columns in the same line. Please do not average over your data beforehand

Conditions coded as factors: If you have a factorial design, that is you are interested in the interplay and different combinations of at least two independent variables, please code them separately: each independent variable should be represented in one column.

All possible confounding factors coded: Apart from the independent variables, there are often other factors which could play a role and which can be included in the statistical analysis. Examples are length and frequency (for words), age, language profiency (L2), error rate, trial number etc. If you do have information on some of them, please try to include them in the file as individual columns.

Random factors coded: for most linguistic designs, this means coding participant number and item number as variables (columns).

While r can handle excel files and spss files, simple text files where columns are separated by tabs are preferred.