Graph Interpretation Project

[Whatshisname]

Hello, my name is Michael.
For my sophomore science project, I've decided to do my own interpretation of the project found here:
https://www.sciencebuddies.org/science- ... Q&from=TSW
Instead of covering the six or seven unique graphs recommended by the project, which require two values each, I've decided to steamline it and focus on the aligned bar graph and the pie graph.
I want to experiment the accuracy of guesses based on a variety of factors- the most important being pie graph or bar graph, (Which is better?) scale, (Can the scale of a graph skew the results?) type of shading and color of shading, (Does grayscale work best? All white? Random Colors? What about varied shading, or as a gradient representing the value?) and finally, what kind of people make more accurate assumptions (Males versus females? Teenagers versus seniors?).

I need help interviewing qualified experts in this field, be it expertise in human psychology or as an expert on visual perception. The only information I need about the interviewee is a first name and a description of their field/expertise/job/experience.
I have 5 questions about visual graph perception overall:
1. Which gender generally has more precise answers? Does gender even matter when looking at graphs? Is the type of graph important to the gender (for example, are men better at pie graphs and women better at bar graphs?
2. Is age a factor in accuracy? Is a specific age group the best, and if so, why?
3. Have there been proven studies about the effects of bar graphs versus pie graphs, or any other kinds of graphs, and how effective they are with different data?
4. Does color or shading in any way affect accuracy? Is having the same color for all sets of data better than having bars/pie slices with different colors?
5. Are there any other studies you can recommend that would further my understanding of graphical perception, or the psychology of data presentation?

I'd like to thank you for reading through all of this, and if there is any way you can help, I would appreciate it. Fortunately I don't need urgent answers, but I hope to have these questions answered as soon as possible, so feel free to recommend me to studies and lab reports. Every little bit helps.

[davidkallman]

Hi Michael,

In partial response to your questions,

Re accuracy: if accuracy is important, it's best to label a point with its numeric value.
Re perception: light colors, yellow and red stand out best It's also good to make one maybe two points per graphic.
Re interviews; my wife is an expert in this field having owned a graphics design firm for almost ten years. I'm not sure what the logistics of an interview will be. I'll have to check with the sciencebuddies folks about the logistics.

[davidkallman]

Hi Michael,

A few items:

1. It's best if we keep all communication on the board, where the all members of the project can view the questions and responses, gain knowledge from them and contribute to the discussion.

2. So, if you have specific questions, please post them here.

3. Re the past posts: my wife, who has a lot of graphics experience, notes:

a. make sure the data has exactly the same meaning in each comparison set you give the testers. For example, if you compare a stacked bar chart to a pie chart, make sure the elements in the stacked bar chart add up to 100%, since the pie chart elements will, by default, add up to 100%

b. Stick to testers who are qualified and can be separated by one factor you want to test, for example, men vs. women or specific age groups.

c. This is a cognitive science project, more about perception than about graphing. Suggestion is to skip the question about shading as a test factor: a bad feature shouldn't be something you test! (Also, using only one color in either a bar or pie chart would be a bad feature.)

d. Note the references in the bibliography section in the project description on the sciencebuddies site. Do they help? Do you need other background material?

[Whatshisname]

1. Yes, the questions were for here. I just meant that when you post your answers, leave me that first name and your expertise. No need to go anywhere else.

2. The questions I have in my first post are fine.

3. Stacked bar graph? ...well, I was going to do individual bars, but now that I think about it, that makes a lot more sense. Yes, having them add up to 100% will work, and the test will be to see how many percentage points they are off from the actual value. Thanks for that suggestion!

I was hoping to have my testers simply fill in their gender and age group (14 and Under, 15-20, 21-30, 31-40, 41-50, 51 and Up). Is there any specific reason why doing both at the same time is a bad idea? It doesn't seem like it would skew the data... not to mention I would be able to use more of my results. Tell me why you think I shouldn't do this.

I believe that if shading is something that could possibly affect the accuracy of answers, its worth testing for. For graphs of all one color, I was planning to split the pie pieces/bars with thin black lines, and the shading would be different for each piece when they were shaded differently. Those aren't graph-ruining aspects, and I want to at least find out if they makes a difference.

4. Eh... Wikipedia pages aren't going to be credible help, and I don't need those semi-unrelated neuroscienece theories confusing the visitors to my project. I think I'll just stick to the facts. I don't need any other sort of background research, I just need the interviews.