Would you like to be better at visualization? Then I want your help!

[sushi]

Well, it's that time of time again. Most of the people currently doing this are due to report in the next few days. I'll update the results in a week or so.

[sushi]

Anyone else care to report back? At present, we only have one for the month. About one in nine of the people who start this finish it.

[Yakumo]

I filed my report.

Short summary - it's not going very well. While I feel I did better in the test, my visualization skills have not improved, I'd even say I was better one month ago. Probably should have spent more time visualizing than doing other exercises.

 

Still I'll continue, see ya in 1 month

[sushi]

Thanks. There aren't many responses, but the ones we've gotten are interesting. I think I made some mistakes the last time I posted the results, so I'm doing everything over from scratch. I should have this posted in the next few days.

[sushi]

All the results are posted in the first post.

 

I've reworked this test. The timed portion is gone, so now it's easier and faster to take this. I've removed that section because it doesn't seem to be doing us any good. You can see the results of that in a graph linked to in the first post.

 

Also, if you look at the results at the bottom of the first post, you'll see that two of our subjects actually got worse after a month of visualization. I think that might be because a person's visualization ability fluctuates throughout the day. For example, your visualization ability might be high after you wake up, then go down as you eat breakfast and watch TV, then you draw for a bit and it goes up again, but then it goes down when you drive to work.

 

Not sure if that's true, but it might explain the results we've gotten so far. So I'd like people to test themselves before and after some brief activities. Test yourself before drawing a picture and after drawing it. Test yourself before your forcing session, and after it. Test yourself before playing chess and after it. This way we can look at all of the tests to get an idea of your progress, but we'll also be able to compare befores to afters to see if certain activities trigger a sharp increase in ability.

 

Also, you can now add up your own results. This way you can compare your befores to your afters yourself, and chart your own progress. I will continue to do this periodically myself, and post the results here, but if you want your results faster, you can add them up yourself. The way I do it is:

 

No image at all: 0

Vague and dim: 1

Moderately clear: 2

Clear and reasonably vivid: 3

Perfectly clear: 4

 

These scores will be listed on the bottom of the test page where you're answering those questions, so you can add everything up while you're on that page.

 

I add up all the answers and that's the score I compare to others. If your before was 36 and your after was 41, you can go after divided by before, or 41/36=1.13. That's a 13% improvement.

[Yakumo]

Well as said before, my main problem with this test is that it tries to measure something which is pretty hard to quantify for an individual and probably incomparable between different individuals. What's a good visualization for me may be bad for you and vice versa. Also just rating yourself is pretty lame, the timed part was the only interesting one.

 

About the graph - I don't get it at all.

What units are plotted on the axes and what do the lines and peaks actually represent? Shouldn't there rather be discrete datapoints for each individual?

 

Being 'Subject 7' (I should change my forum name to this), I hate to say but it's no artifact, I actually feel I'm doing worse at visualizing than a month ago. My timing has been fairly accurate, but for the questionnaire, it's hard to choose between 1 and 2 as my results are somewhere in between - but this month they were closer to 1 than last month.

[sushi]

I wonder, should I revert this to the way that it was, or maybe run both tests side by side? Maybe I should keep the timed section separate from Marks' questionnaire?

 

Sorry I didn't explain the graph so well.

 

The horizontal axis is just a different point for each person. Now that you point it out, this was not the best way to graph this information.

 

The blue line is the Marks score. I explained how that's scored in my last post. So it ranges from a lowest possible score of 0 to a highest possible score of 64. I reordered the people so that the lowest scores (16) are on the left, and the highest scores (61) are on the right. About half of the scores fall within five points of the 40s.

 

The other three lines are x/(y+1), where x is the time (in seconds), and y is the number of mistakes. The magenta line is mistakes only on the letter square section, the yellow line is mistakes only on the shapes section, and the cyan line is mistakes on the full timed section.

 

I know a person's Marks' score is very subjective. One person might imagine themselves a better visualizer than another person imagines themselves, when they really aren't. I was hoping that with enough responses, those would balance each other out.

 

Anyway, I'd like to see this be useful somehow, but I clearly don't know how I should be doing this. Does anyone have any suggestions?

[Yakumo]

Ah, I see...

That makes it a bit clearer. So you have 57 individual test results? I thought it was way less. If that's really the case we should be able to do some decent statistics with this dataset.

 

Well I 'd say first you need to get rid of the lines, you don't have continuous data on the x-axis but discrete data-points. Use bars or dots for each individual.

 

Second I'd use one scale for the left y-axis (mistakes) and another one for the right one (reported scores). This way you can spread the scores for better display. Right now compared to the upper values of 700 for the mistakes, the score value range between 0-64 shows almost no difference between the individuals. You don't need units on the x-axis.

 

Personally I'd rather use (y+1)/x as errors per time but you need to test what works out best in the end. You could even include errors only and time only. Maybe it yields interesting results, maybe it doesn't. Just experiment with different variables, you don't need to put everything into one graph.

 

Finally, if you want to see correlation I'd do 3 scatter plots, one with each error form (letters, shapes, both) against the reported score. You can also test if time plays a role or have it omitted to measure mistakes only, that would be interesting. Any decent program, even excel, should give you a correlation coefficient for the two variables - a value between +1 and −1, where 1 is total positive correlation, 0 is no correlation, and −1 is total negative correlation. That's way more informative than the current graph.

 

And - most important rule when it comes to graphs:

Be sure to label your axes and give a short description about what's going on.

No exceptions.

 

I'm on holiday now 'til Jan 6th but I can give it a look afterwards if you like. No expert when it comes to statistics but I do know some basics and work with this stuff every now and then. I also have the software to produce neat graphs and do advanced calculations if that's needed.

 

Anyway, thx for all the work, I'd not give up on the subject yet. Maybe we still find some correlation but if we do not, I wouldn't be surprised either.

[sushi]

Ah, I see...

That makes it a bit clearer. So you have 57 individual test results? I thought it was way less. If that's really the case we should be able to do some decent statistics with this dataset.

 

Let's say the test has been taken 57 times. Seven of those are people taking the test for the second time, and one of this is someone taking it for the third time. So most of these are useless for us comparing progress, but they should give us a correlation between the two parts of the test.

 

I'll have to put further graphing off until tomorrow, as I only have a few minutes left before I leave work. I may need to take you up on the use of your software. I'm using gnumeric, and I'm not sure what graphing options I have available to me. I'll drop some more graphs in here tomorrow, and we'll see what we can see. Though I'm not expecting to see anything.

[Yakumo]

That should be fine regarding the dataset.

I'll probably not be able to take a look at your new graphs until I'm back, but I definitely will do so in the 2nd week of January.

There truly may not be any correlation between the ability to visualize and the ability to do well in these tests but that's what we are trying to find out here.