Information Visualization: Assignment 6 - Critiquing Multivariate Visualization System 2 | CS 7450, Assignments of Computer Science

Download Information Visualization: Assignment 6 - Critiquing Multivariate Visualization System 2 | CS 7450 and more Assignments Computer Science in PDF only on Docsity! CS7450 Homework 6: Critiquing Multivariate Visualization Systems 2 Urs Bischoff I have organized my report as follows. First, I show the example sets and the tasks for the set that I have chosen. Then I show my findings for these tasks for each of the tools. Then, I describe the strengths, weaknesses and an interesting finding for each of the tools. Finally, I compare the two tools based on a task taxonomy. Examine the sample data sets I have chosen the cars and the cereals data set. Hypotheses/Tasks/Questions Do cars with more cylinders have more horsepower? Which European car has the highest MPG? American cars have more cylinders than European or Japanese cars. What is the distribution of MPG? I am interested in food that has low fat but high protein and carbohydrates. How much fat does Kellogg’s Special K contain? What is the distribution of sugars? What is the difference between Corn Chex and Bran Chex? Load and examine data sets in the system (InfoZoom) Do cars with more cylinders have more horsepower? I used the compressed mode. First, I sorted the attribute cylinder. I selected both 6 and 8 cylinder. Then, I sorted the horsepower attribute. Because 6 and 8 cylinder cars are colored black (I selected them before), it’s easy to see that they tend to have more horsepower than cars with fewer cylinders. But also the horizontal lines (that visually show numerical values) supported my assumptions. In Figure 1 you can see that the black colored cylinder fields tend to be on the right side. Cars are sorted so that those with more horsepower are on the right side. If there were more than two variables, it would useful to use another available visualization technique (e.g. scatterplot). I wondered if there’s a “linear” relation between cylinders and horsepower; i.e., is the amount of horsepower per cylinder for all cars about the same. This would mean that by doubling the number of cylinders, I also double the horsepower. I doubt that this is true. I used derived attributes (attributes that are based on a function of existing attributes). I couldn’t identify a correlation. However, the car (mazda rx-4) that has the best ratio (horsepower/cylinder) is a car with 3 cylinders; the car that has the worst ratio is an eight cylinder car (cutlass salon brougham from oldsmobile). To detect the lowest and the highest value, the arrows on the left of the attribute names were very useful. They showed all possible values and the frequency in a list. I selected the best ratios and zoomed in. I could see that there were 3-, 4- and 8-cylinder cars in this set. Figure 1: Cylinders and Horsepower Which European car has the highest MPG? I found that the Overview Mode was the best mode to answer this question, because the values of each attribute are already ordered separately. I started by zooming in on European cars (Figure 2 shows the screen). Then, I selected the rightmost field in the MPG row (if there are too many fields it is difficult to select the rightmost field; I prefer to use the list that is opened when I click on the arrow left of the attribute name). I found out that VW’s rabbit c is the best here. American cars have more cylinders than European or Japanese cars. I used the compressed mode. First, I sorted by cylinders than by origin. I could clearly see a trend that American cars have more cylinders. The horizontal lines in cells are very useful to compare numerical values (e.g. number of cylinders). I was then also interested if they also have more cylinders in average. I used the derived attributes functionality and the compressed mode to compute the average number of cylinders. Overall, the average number of cylinders is 5.48; for American cars it is 6.28, for Europeans 4.15 and for Japanese 4.1. So, American cars also have the highest average. een ed Ble Edit View Adjust Help SHR Soo 7 ae mu erL Ba 13 12 ci 10 Number of products ~ yo Sugar Values 600x400 4 Figure 4: Distribution of sugars il Cereal il Manufacturer Type Calories ld tld ld Protein id Fat iid Sodium Fiber iid il Carbohydrates tk = Sugars Shelf Potassium ld tld ld Vitamins = weight = Cups Bran Chex Corn Chex R c 30 110 2 2 1.00 o.oo 200 280 4 a 16 22 6 a 1 1 125 25 25 25 1 1 o67 1 Figure 5: Comparison  Load and examine data sets in the system (EZChooser) Do cars with more cylinders have more horsepower? I selected the 6 and 8 cylinder cars. All cars in this set are also marked in the other feature rows. Thus, I could see where those elements are in the horsepower row. Unfortunately, the horsepower row isn’t completely sorted, i.e. the cars with a lot of horsepower are somewhere in the middle of the row. But I can still see that more or less all cars with a lot of horsepower are selected. Figure 6 shows the result. I also tried to answer this question by first selecting those cars that have a lot of horsepower. However, it is very difficult, because the buttons that should be pressed are very small and because the horsepower row isn’t sorted. Figure 6: Cylinders and Horsepower Which European car has the highest MPG? I first clicked on the button that selected all European cars. Then, I used the filter to narrow down the displayed cars to European cars. Fortunately, the MPG row was sorted; so I only had to select the rightmost element. I could read the result in the bottom window: it is the rabbit c (diesel). American cars have more cylinders than European or Japanese cars. I approached this problem from two different sides. First, I selected all American cars. Because the cylinder row was sorted, I could see that most of the selected cars (American cars) are on the right side, which means that they have more of cylinders. But this does only say that American cars have a lot of cylinders. But it doesn’t say that they have more cylinders than European or Japanese cars. That’s why I also approached the problem from the other side: I selected all 6 and 8 cylinder cars. I could clearly see that most of these cars were American cars. What is the distribution of MPG? I wasn’t able to satisfyingly solve this task. Unfortunately, there aren’t any visual indicators that visualize the numerical values. I cannot even read the values on small buttons. I tried to get a feeling about the distribution by moving my mouse over the row. If you move your mouse on a button, it displays the value of the button. However, I didn’t get a real feeling about the distribution. I am interested in food that has low fat but high protein and carbohydrates. I used a filter to only display the products that have a fat value of 0 or 1. Then, I used the three different colors to mark the cereals with a high carbohydrate and the ones with high protein values. By playing with these colors, I came to the conclusion that Kellogg’s Special K is the best product for my needs. Figure 7 shows a screenshot: only low fat cereals are shown. I colored high protein cereals with different colors. The “green” product had the best carbohydrate value. It was Kellogg’s Special K. Figure 7: Low fat and high protein cereals are colored How much fat does Kellogg’s Special K contain? I narrowed down the selection by only displaying Kellogg’s products (I clicked on the K-button in the manufacturer row). Then, I could identify the name Special K in the cereal name row. I used the filter to only display Special K. It has a fat value of 0. What is the distribution of sugars? As I’ve already mentioned earlier, EZchooser is not suitable for this task. Here, it’s easier to get a rough feeling about the distribution than before (distribution of MPG), because there’s only a small number of different sugar values. Unfortunately, EZchooser builds buckets to collect values. What is the difference between Corn Chex and Bran Chex? This task could also be very difficult, because the resolution of the screen is too low to display the full names of all products on the buttons. The names are sorted, but it’s still difficult to identify a product. However, I was lucky because both products appeared on the list in the bottom window. Only a selection of all products is displayed there. I think it’s not possible to also display the names of the other products in that window. However, both products were displayed. Thus, it was easy to select those two and use a filter so that only those two are displayed. Figure 8 shows the resulting screen. One has to be very careful! Because each row is sorted separately, not all values in one column belong to the same product. I used two different colors to make the interpretation easier; otherwise the visualization could be very misleading. Figure 8: Corn Chex and Bran Chex Critique of the Tools In this subsection I try to critique the two tools. Some of the comments have already been mentioned in the previous sections. I concentrate on the strengths and weaknesses of the tools. (A comparison of the tools with respect to user tasks can be found in the next section.). Some of the comments are directly related to the previous tasks.