[PYTHON] Build a comfortable psychological experiment / analysis environment with PsychoPy + Jupyter Notebook

Preface

• For those who are only Coder and do not use Builder very much.
• Alternatively, for those who want to experiment with code but have trouble preparing the development environment.
• The installation procedure is Mac compliant, but Windows is similar.

PsychoPy as a library is very complete and much easier to use than Pyglet, Pygame and Kivy, while PsychoPy as an IDE is subtle with weak completion and debugging.

I used to edit scripts in Vim, but recently I've noticed that PsychoPy works very well with Jupyter.

So, this time I will explain how to use PsychoPy with Jupyter.

Before ..., let's first introduce how useful it is.

gallery

Of course it can be started.

Want to Hello World? ... then you can add it later.

Oops, I wanted to make black letters on a white background. Close the page with win.close () ... What was the color setting of visual.Window?

Remake and re-present.

... you can create experiments interactively like this.

Environment creation

Installation of Anaconda

Use ** Anaconda ** for later analysis. Click here to install.

• Anaconda for Windows
• Anaconda for Mac

There are 2 series and 3 series, but here we will include 3 series [^ 1]. The reason is as follows.

• Even if you install 3 series, you can use 2 series Python environment without any problem with the settings described later.
• There are almost no libraries other than PsychoPy that do not support 3 series.
• The PsychoPy is also trying to move to 3 series
• Support for 2 systems will be discontinued in 2020 [https://pythonclock.org/)

Also, I think it is good to install Miniconda instead of Anaconda. The reason is as follows.

• Anaconda consumes capacity (uses 2 digit GB without hesitation)
• The introduction of Anaconda has some sober side effects such as hiding the curl path.

However, be aware that if you use Miniconda, you will need to include the libraries (numpy, pandas, etc.) yourself.

Operation check of Jupyter Notebook

After installing Anaconda, first make sure that Jupyter Notebook starts. Open a terminal (Anaconda Prompt for Win) and do the following:

jupyter notebook

If the browser starts up and localhost: 8888 is displayed automatically, it is successful.

Please do not delete the terminal or Anaconda Prompt opened here while using Jupyter Notebook. If you turn it off, the server will stop. On the contrary, if you do not use it anymore, you can close it.

Install PsychoPy

Open Anaconda Prompt (Windows users should run as administrator) and do the following:

pip install psychopy

After installation, try the following from your notebook:

from psychopy import visual

win = visual.Window()

A window should open.

In case of Windows, NotImplementedError may be displayed. In that case, open Anaconda Prompt again with administrator privileges and open it.

pip install pyglet==1.2

(I'm lowering the version because it doesn't work with version 1.3, which comes in by default).

Supplement

behavior of visual.Window

If you try to drag and drop a screen created with fullscr = False, it may become unresponsive. There is no particular problem, and the stimulus is presented properly, so it is okay to proceed with the experiment creation as it is.

Execution speed measurement

With% time or% timeit.
Please when the stimulus presentation is slow.
When set to% 2, the execution time of the entire block is measured.

Cooperation with R

If you want to use R with Jupyter Notebook,

https://irkernel.github.io/installation/

Let's put R referring to. Open R

install.packages(c('repr', 'IRdisplay', 'evaluate', 'crayon', 'pbdZMQ', 'devtools', 'uuid', 'digest'))
devtools::install_github('IRkernel/IRkernel')
IRkernel::installspec()

If you execute ** paying attention to the following **, it will be automatically linked with Jupyter Notebook.

[Caution (Win)]

• R must be started ** with administrator privileges **
• Install Rtools in advance from https://cran.ism.ac.jp/
  (In windows, devtools should not be included in install.package)

If ʻIRKernel :: installspec () `is moss, the following measures may be effective.

• ʻIRKernel :: installspec () `only runs on R launched from Anaconda Prompt
• If you cannot start R from Anaconda Prompt, pass the path to R in the environment variable "Path"
• By default, the path to R is C: \ Program Files \ R \ R-X.X.X \ bin \ R.exe etc.

[Caution (Mac)]

• R must be launched ** from the terminal **

[^ 1]: This article was also compatible with both 2 series and 3 series until January 2018. Since there is almost no merit to install the 2nd system now, I modified it to introduce only the introduction method of the 3rd system.