dmcanzi@watdcsu.UUCP (David Canzi) (07/10/85)
I found the following interesting uses (abuses?) of statistics in last night's copy of the local newspaper. Three pie charts were presented, showing how time is divided into free time, paid work, and unpaid work (usually housework), for three groups of people: housewives, employed men, and unemployed men. The numbers in each wedge of the pie charts were probably percentages, since they add up to values near 100. The sizes of the wedges in the pie charts did not accurately reflect the values of the numbers. The following table shows who has how much free time, according to the numbers, and according to measurement of the pie chart wedges: housewives employed men employed women by the numbers 33% 34% 24% by wedge sizes 33% 46% 17% Ie. the pie charts appear to be deliberately distorted to make the womens' lot seem worse than it really is, and the mens' lot look better. A bar graph was presented showing unemployment rates for men and women in various countries. In most countries, except the USA and Japan, unemployment was higher for women (in Italy and France, extremely so). Lest the reader be fooled by the nice figures for the USA and Japan, the author of the article included a comment to the effect that "In Europe only 42% of women looking for work are registered as unemployed." THE CORRESPONDING PERCENTAGE FOR UNEMPLOYED MEN WAS NOT PROVIDED, and the European figure for "unofficially" unemployed women may be larger than the American and Japanese figures for the same. Another chart showed that the percentage of the work force consisting of women has increased only slightly (from 31% to 35%) since 1950, and not at all (stuck at 35%) over the last decade. The cute diagram illustrates 35% with a picture showing 12 men and 4 women for 1975, 15 men and 5 women for 1985. Ie. the picture looks worse than the numbers, just as in the pie charts. (No interpretation of the possible causes of this 35% ratio was offered.) Another chart was presented showing that women in Africa do somewhat more than a fair share of agricultural labour, and 95% of domestic work. No indication is given as to what percentage of the total amount of work performed is agricultural or domestic, and how much is the man's share of other kinds of work. Thus, the figures given almost certainly exaggerate the degree to which women are being overworked in Africa. A litle bit of advice to you all: 1) When you see statistics being quoted, check for ways in which they might have been exaggerated to support some political viewpoint. 2) When you quote statistics yourself, make sure they are reasonably complete, and really support the viewpoint that you claim they support, 'cause there will always be some prick like me, waiting to point out the flaws. If you fudge the stats and get caught at it, you ruin your own credibility. -- David Canzi "With the exception of victimless crimes (which need not concern us here) every single crime committed in this nation of ours involves a victim." -- San Francisco Chronicle
dmcanzi@watdcsu.UUCP (David Canzi) (07/10/85)
I just noticed another interesting feature of the diagram that showed women as 35% of the work force... the diagram is a picture of a single-file line of men at the back and women at the front. Not only did the person who drew the diagram draw 25% of the figures as women to supposedly represent the 35% of the work force that women comprise, but the women in the drawing were drawn CLOSER TOGETHER than the men, making them look like an even smaller portion of the total. -- David Canzi "With the exception of victimless crimes (which need not concern us here) every single crime committed in this nation of ours involves a victim." -- San Francisco Chronicle
sophie@mnetor.UUCP (Sophie Quigley) (07/11/85)
> I just noticed another interesting feature of the diagram that showed > women as 35% of the work force... the diagram is a picture of a > single-file line of men at the back and women at the front. Not only > did the person who drew the diagram draw 25% of the figures as women to > supposedly represent the 35% of the work force that women comprise, but > the women in the drawing were drawn CLOSER TOGETHER than the men, > making them look like an even smaller portion of the total. > -- > David Canzi Aren't you reading a bit too much into this? maybe women stand closer together than men or something like that. -- Sophie Quigley {allegra|decvax|ihnp4|linus|watmath}!utzoo!mnetor!sophie
arnold@ucsfcgl.UUCP (Ken Arnold%CGL) (07/13/85)
[ yum yum ] In article <1520@watdcsu.UUCP> dmcanzi@watdcsu.UUCP (David Canzi) writes: >I just noticed another interesting feature of the diagram that showed >women as 35% of the work force... the diagram is a picture of a >single-file line of men at the back and women at the front. Not only >did the person who drew the diagram draw 25% of the figures as women to >supposedly represent the 35% of the work force that women comprise, but >the women in the drawing were drawn CLOSER TOGETHER than the men, >making them look like an even smaller portion of the total. >-- >David Canzi (This is a followup to a long article on abuse of statistics) As has been occasionally mentioned (far too rarely to suit me) not only can statistics be manipulated to favor given points of view, charts are a great source of abuse. The best way to educate yourself about what to be wary of is to buy and read "How to Lie With Statistics", by Darrell Huff, copyright 1954, published by W. W. Norton & Company, Inc., New York. I would suspect it is quite a bit out of print, but it is commonly available in used book stores. You will never find yourself at a loss again when someone spouts statistics which make no sense, but you can't quite figure out why. As a matter of fact, you will see many statistics and charts which you thought make sense don't. Ken Arnold
dmcanzi@watdcsu.UUCP (David Canzi) (07/16/85)
In article <1266@mnetor.UUCP> sophie@mnetor.UUCP (Sophie Quigley) writes: >> Not only >> did the person who drew the diagram draw 25% of the figures as women to >> supposedly represent the 35% of the work force that women comprise, but >> the women in the drawing were drawn CLOSER TOGETHER than the men, >> making them look like an even smaller portion of the total. >> -- >> David Canzi > >Aren't you reading a bit too much into this? maybe women stand closer >together than men or something like that. >-- >Sophie Quigley The diagram was drawn as a single-file line of people standing in a file drawer facing the viewer (cute, nuh?), with women at the front and men at the back. The top of each man's head was drawn at about the level of the chin of the man behind, while the top of each woman's head was at about nose-level of the woman behind. The top of the last woman's head was chin-level on the man behind. The first woman was visible from the chest up. If the first woman's chest is included when measuring the diagram, the women make up roughly 1/4 of the measured length of the diagram. If the measurement is taken from the first woman's chin (which is my natural impulse), then the diagram is exaggerated. It could be that most people's visual impression of the diagram is the impression that results from including the first woman's chest. It could then be supposed that the artist drew the women closer together in order to make up for the length added to the diagram by the first woman's chest, but an artist with sufficient smarts to do that would also have known better than to represent 35% of the working force with only 25% of the human figures in the drawing. For whatever reason (perhaps an illusion resulting from perspective), the visual impression I got from looking at the diagram was that women were even less than 25% of the diagram. Some of the problems I pointed out in the original article could be innocent errors, but in the case of the pie charts and the "file drawers", I can't help but think that the people producing the diagrams weren't satisfied with the truth and felt a need to "enhance" it. According to the newspaper article, the pie charts are from the Worldwatch Institute, and the "file drawers" are from the ILO. I'm not very familiar with these organizations... are they perhaps in the habit of producing misleading statistical diagrams? -- David Canzi "Adequacy -- is it enough?"
crs@lanl.ARPA (07/16/85)
Another common stratagem is the use of breaks to emphasize small (often insignificant) differences. For example: | 99.99 --------# | # | # | # | # 99.98 --------#-------@ | # @ | # @ | # @ | # @ ~ ~ ~ ~ ~ ~ | # @ | # @ 0.00 ---------------------- X Y -- All opinions are mine alone... Charlie Sorsby ...!{cmcl2,ihnp4,...}!lanl!crs crs@lanl.arpa
herbie@watdcsu.UUCP (Herb Chong [DCS]) (07/26/85)
i had a look at the newspaper article that David was refering to when he started this discussion. in the pie charts, a wedge labelled 40% actually occupied about 47% of the area and another part labelled 24% occupied closer to 20% of the area. lest you think that the difference isn't all that great, i suggest you draw a chart and divide it up accurately using a protractor. visually, the difference is tremendous. another chart had a slice labelled 49%, but occupied only 40% of the area, and 34% occupied about 45% of the area. what is even more interesting is that in none of the pie charts do the figures add up to 100%. in the corner of the diagram, the source cited is ILO. i'm not sure if this means that the actual pie charts were reproduced or they were just the source of the numbers. whatever the source, the figures are grossly misleading and are also the thing that people remember the easiest. the bar chart representing proportion of women in the workforce is just as bad. for one thing, the ratio of the number of male figures to female figures is much larger than the percentage figures on the labels. for another, the male figures are physically larger then the female figures, although one supposes that one male figure equals one female figure in the chart. the worst thing, and slightly more subtle, but more damaging is that the file cabinets in which these figures are placed are drawn in prespective while the figures are identical in size from front to back. this makes the male figures in the back appear much larger than the female figures in front. it is unknown whether the person who drew these figures had these effects in mind when they were designed, but they definitely are slanted toward a viewpoint that women are overwhelmingly overworked and discriminated against. the numbers in the figures support this view, but not in the visual proportions implied. i don't think it is coincidence that the newspaper article was about how far women still had to go in fighting discrimination. i think that the misrepresentation made by the diagrams certainly influences opinions, but can make some of the claims in the article less believeable. there was a book several years ago that i read that was titled something like "How to Lie With Figures" or something like that. it is entirely devoted to the techniques such as described above so that people reading papers and such would not be caught by visually misrepresenting figures. i am not judging content of the articles but the misleading method of visual representation of numbers clearly indicated in the diagrams. Herb Chong... I'm user-friendly -- I don't byte, I nybble.... UUCP: {decvax|utzoo|ihnp4|allegra|clyde}!watmath!water!watdcsu!herbie CSNET: herbie%watdcsu@waterloo.csnet ARPA: herbie%watdcsu%waterloo.csnet@csnet-relay.arpa NETNORTH, BITNET, EARN: herbie@watdcs, herbie@watdcsu