Javascript Gaussian function

Online experiments
1

URL of experiment: Pavlovia

Description of the problem:
I am looking to set my experiment up to run online and my x coordinate for my stimulus position values are set in a code component.

Using python code I set the x position for my two stimuli to come from two gaussians where the value of the target becomes the mean of the cue:

target_coin_position = random.gauss (mu, priorsigma)

cue_coin_position = random.gass(target_coin_position, cuesigma)

Is there an equivalent method of getting values from a Gaussian distribution in java script?

I did some research online and found randomGaussian(mu, sigma)

So I tried the java script code:

target_coin_position = randomGaussian(mu, priorsigma);

cue_coin_position = randomGaussian(target_coin_position, cue_sigma);

For this my error was:

randomGausian() is not defined- see screen shot image

I also looked into Random.nextGaussian() but I think that was for Java not Java script.

If I want this to work online is there any adjustments I could make or is this function not available to PsychoJS currently?

Could I do this process manually in java script by creating function that takes the mean (mu) and SD (sigma) as parameters and creates a normal distribution from those parameters?

Screenshot 2020-11-02 at 10.18.16
Screenshot 2020-11-02 at 10.18.161024Ă—572 29.8 KB

2

Hi Lucy,

There is no built-in function available, but there are some libraries on the web that you could use. Some aren’t very solid, but when I needed a quantile function for a d-prime (see dprime repo), I ran some tests and found the function in this repo to be reliable. Above, I linked directly to the function you need.

Now, I would recommend testing it first, for instance by generating a large set of numbers with a Gaussian distribution and draw a histogram with a normal curve overlaying it. If you run the script below in the browser console, you can get these numbers.

// Function from the gaussian repo, with a slight adjustment
function generateGaussian(mean,std){
  var _2PI = Math.PI * 2;
  var u1 = Math.random();
  var u2 = Math.random();
  
  var z0 = Math.sqrt(-2.0 * Math.log(u1)) * Math.cos(_2PI * u2);
  var z1 = Math.sqrt(-2.0 * Math.log(u1)) * Math.sin(_2PI * u2);

  return z0 * std + mean;
}
// Generate 10,000 Gaussian random numbers
result = [];
for (var i = 0; i < 10000; i++) {
  result.push(generateGaussian(0, 1));
}

Best, Thomas

3

I ran the test, using this function in R to plot a Histogram with an normal curve overlaid on top of it. Seems legit :slight_smile:

image
image796Ă—563 4.66 KB

4

Thank you for getting back to me so quickly Thomas. It looks fantastic.

1 Like
5

Hi again Thomas,

I had a go with this and added the function to my code component where it is required. However I think the next problem I am having (I think) is importing the libraries at the very beginning (see .js file in R studio) because the experiment does initialise. I then type in the participant details then google chrome crashes and I have to reload (I’ve explored the help chrome offers for this deleting cache, updating and restarting etc.)

Screenshot 2020-11-03 at 11.30.05
Screenshot 2020-11-03 at 11.30.051184Ă—996 147 KB

When I open the developer tools I get these errors (see below) which I am guessing relates to the fact it cannot import the sound libraries?

Screenshot 2020-11-03 at 14.49.41
Screenshot 2020-11-03 at 14.49.411160Ă—934 156 KB

6

Never knew that RStudio supports syntax highlighting for JavaScript. Interesting!

The warnings RStudio gives on those lines nothing to worry about; it’s just a bit more modern style of JS than RStudio is familiar with.

When I fork and run your experiment, I notice my browser freezes. When I stop it, I get the message “script terminated by timeout” at a line inside of the loop below. My guess is that this while loop takes very long (maybe never ends).

  for (var j = 0, _pj_a = ntrials; (j < _pj_a); j += 1) {
      target_coin_position = result.push(generateGaussian(mu, priorsigma));
      cue_coin_position = result.push(generateGaussian(target_coin_position, cuesigma));
      while (((((target_coin_position < minposition) || (target_coin_position > maxposition)) || (cue_coin_position < minposition)) || (cue_coin_position > maxposition))) {
          target_coin_position = result.push(generateGaussian(mu, priorsigma));
          cue_coin_position = result.push(generateGaussian(target_coin_position, cuesigma));
          
      
      }
      target_list.append(target_coin_position);
      cue_list.append(cue_coin_position);
  }
7

Okay thank you I will try running it without the while loop (it only provides a constrain for the x axis value) and then see if I can adjust (add break or similar).

1 Like