Blog
Statistics for Sustainable Development > Blog > Statistical modelling with R: Training Tips on how to keep Participants Engaged
Statistical modelling with R: Training Tips on how to keep Participants Engaged
Since 2014, I have been providing Research Methods Support
to McKnight Foundation funded projects in Eastern Africa. Currently, I am based
in Uganda and sit within Makerere University (Kampala), where 40% of my time is
allocated to work with students within the University. Last summer, I was
requested to take on an additional task that involved taking PhD students
(Agriculture and Rural Innovations) through practical sessions on statistical
modelling with R. This struck me as both an exciting and challenging task, and
I set to work by discussing with my supervisor and colleagues on how best to
approach the training.
The view from outside Makerere
University, Kampala in Uganda
Since my requests for a data set to use during the session
did not yield any results, I decided to use some of the experimental data sets
from Stats4SD that are used for introducing project teams to R. During the
first training session, it was clear that the students (or the course they were
undertaking) were mostly prepared for using survey study designs, and that most
of the experimental jargon was new to them. By the end of this initial session,
it was clear that working with experimental data was not the way to go and I needed
to change strategy. The next option was to request a data set from the students
that they could all relate to. Unfortunately, the students had not yet
collected any data and we could therefore not pursue this option. So finally, I
went for the third option that involved using an old US Employee data set. With
this, the students could easily relate to what they had covered in the theoretical
part of the class - and were able to actively participate.
During the week of training, the following topics were
covered:
- Importing
the data set from SPSS
- Checking
data structure
- Generating
summaries by groups
- Simple
ggplots
- Measures
of central tendency
- Correlation
- Chi-sq
- Tests
for normality and homogeneity of variance
- T-tests
and OLS.
- Logistic
regression