Delivery
The current PSYU914 has a delivery mechanism with 4 parts:
- Online video recordings covering
- all matters conceptual
- Workbooks that cover:
- demonstrations of concepts
- practicalities of data analysis
- Tutorials focussed on the workbooks:
- emphasise the conceptual elements
- assist with the practicalities
- Live lectures covering
- wider context
Assuming that the basic curricular content remains unchanged, then there are ways in which delivery can be improved. Here are some thoughts.
Options
Reduce overlap.
Tutorials are the least necessary (not to say that they are not needed in some form). Many/most students manage the workbooks on their own – for them attendance at tutorials is unnecessary. It might be possible to replace the tutorials with one or two drop-in sessions per week. Each drop-in session could have one or two paid TAs plus as many volunteer tutors as possible. Students come with discrete questions/issues that might range from “I can’t get BrawStats to work” to “I can’t get Jamovi to do the analysis I want.”
To support this, the first half of the live lectures could work through the Workbook of the week maybe with a short Q&A pause before looking at some context for the material of that workbook. Lectures are recorded so students could use this for their own pass through the workbook. The ideal pattern might be:
- student attempts workbook
- student attends lecture and sees workbook done
- student still unsure comes to a drop-in
Benefits/Costs
The major benefit here is administrative. No need to organise tutors and students into tutorials. “No shows” are not a problem.
It also places more of the responsibility on to the student – there’s no longer an option to turn up to the tutorial and just let others lead you through it.
However, the tutorials have a social element as well. Students see each other struggling with stuff and gain re-assurance that hey are not different from others.
And the dept won’t like the reduction in contact hours that this would create.
Dreaming
Imagine if each staff member set up a weekly Personal Tutor meeting for their students at which 2nd year students worked through the workbooks supported by the staff member plus 3rd/4th year students.
This has the major benefit of involving all staff in what would have to be a common approach to stats.
Software
Currently the module uses 2 stats softwares plus excel in a supporting role:
- Jamovi is designated as the departmental analysis tool – just now. The intention is that this is what students will use for all subsequent data analysis.
- BrawStats is a data simulation tool focussed on demonstrating the conceptual side of statistics. This requires installation of R and RStudio.
An important interaction between the two is that data can be generated in BrawStats and then passed to Jamovi. BrawStats does the correct analysis automatically and students can then try to replicate the same analysis in Jamovi, knowing what numbers they should be seeing.
It does seem that the load involved in supporting the installation of this software is high – not least because RStudio is subject to frequent change. Both are available online for users unable to install the software. In practice this ought to apply only to Chromebook/Android/iPad users.
Options
Simplify the software requirements.
Easier installation.
One simplification would be to insist on students using just one of these two routes:
- Apps Anywhere
Available to all on University computers and students’ own Windows systems. - Downloaded Installations
Available to Windows and MacOS users.
The download files could be hosted on Canvas and be chosen to reflect the same version as Apps Anywhere.
Less installation.
Either of these would work:
- Jamovi only.
In terms of installation problems this appears to be very straight forward. No reported problems with installation.
This route would require one of:
1. drop BrawStats and all use of BrawStats
2. drop all student use of BrawStats but continue to use it for demos in live lectures/tutorials
3. convert BrawStats to a Jamovi add-on
I have a skeletal version of BrawStats working in that way - R Only
Installing R & RStudio has not been straight forward. However, Information Services may have a remit here that can be used so that students try for themselves and when they hit problems they go to the IT Helpdesk.
This route would require all of:
1. beg, borrow or steal a set of R scripts for about 20 analyses
2. rewriting all module materials to use those scripts in place of Jamovi
The route would require one of:
1. align R script outputs with BrawStats (or vice-versa)
2. drop BrawStats and all use of BrawStats
3. drop all student use of BrawStats but continue to use it for demos in live lectures/tutorials
Benefits/costs
Both of these suggestions has the major benefit of reducing demand on staff, especially at the beginning of term. It is also desirable for students that they have a quick and trouble-free start.
The explicit cost is that students using Chromebooks/Android/iPad will have to use university computers. Of these, Chromebook is the only one to worry about as the others are not realistic for degree level study anyway.
The `Less Installation option would involve a degree of revision of module materials. Workbooks, in particular would need rethinking.
Student Learning
Students do not retain skills that they don’t use although they probably relearn them much quicker. The current module was designed around the expectation that students would learn the logic all parts of this process:
- Recognize which aspects of the hypothesis and the form of the data determine the specific statistical test to conduct.
- Identify the appropriate test, perhaps with a reference guide.
- Be confident in finding instructions on how to do the test.
- Safely interpret the results from the test.
- Recommend/design follow-up studies to further explore the hypothesis
The statistical tests in question for this module are all tests with 1 or 2 IVs and 1 DV. IVs can be Interval or Categorical with unlimited number of cases. DV can be Interval or Categorical with just 2 cases. Analysis covers main effects, interactions and covariation (eg mediation/moderation).
It was never intended that students would memorise any part of the practicalities. It was thought better that they would have the ability to find out practicalities for themselves when needed because they knew what they were looking for.
For example, I can’t quite claim to be able to state all the steps involved in watching Lawrence of Arabia on Netflix. But I am quite confident that I could find out how to do it.
The materials downplay two old chestnuts that some are fond of. First, we do not bother students about the (uninformed) issue of non-normal data. There is literally only one circumstance where non-normal data actually matters – where the residuals in a data set are sampled from a Cauchy distribution. I doubt this has ever happened in Psychology. Second, although we introduce ordinal variables, we downplay their importance. The principled reason for this is that they are usually used to deal with non-normal data and we see no issue there. A less principled reason is that Jamovi also doesn’t.
What Students have learned
It will be seen that the 5 steps above do not include “knowing how to do a t-test”. The module produces students how have done those 5 steps repeatedly (as a minimum ~20 times). But if they haven’t encapsulated that learning as “what I know about a t-test” then it won’t be accessed by asking them what they know about t-tests.
If it is to be required of students that they explicitly can say how to do a t-test, then it must also be decided on which software. That takes us back to maybe the software issue.
The risk here is that students can end up learn a procedure (a set of rituals) that they perform without fully understanding what the procedure concerns or being able to move that ritual onto different circumstances (such as data layout or software).
And this is the real tension. It is the procedural knowledge that works best for specific research situations, but the understanding that is the better education.
Options
I see three possible changes if the present situation is judged undesirable by all.
- Use the existing Canvas “Statistics for Psychology” Module much more fully.
This would mean paying someone to bring it up to date and then, ideally, for each module on Canvas to maintain active links to the relevant parts of it. - Spread formal teaching of stats through more of the degree programme.
This would mean giving each module specific stats-content and arranging for that to be taught in a common way.
Coherent content and common teaching method. - Replace the current 914 by a module that more explicitly teaches statistical procedures.
Benefits/costs
The last is, superficially, the least costly from where we are. Experience of 914 over many decades suggests that it would not work – and that the issue is that you can’t teach procedures hermetically sealed away from where they must be used. 914 already has plenty of practical work in it – the issue is in how that experience is used in subsequent modules.
Of the other two, the second is ideal but very costly. The first was set up in 2019 exactly to deal with this issue but needs to be used by staff as well as the students they are teaching. My guess is that most module co-ordinators have forgotten it exists – Lizzie spoke with each module co-ordinator at the time so awareness was good at that time.