4 ms·

Note that it does not ever say that the difference between upper class women and upper class men was statistically significant. > The differences in callback r

Note that it does not ever say that the difference between upper class women and upper class men was statistically significant. > The differences in callback rates for higher-class women, lower-class men and lower-class women were not statistically significant, but higher-class men received significantly more callbacks than all other categories It is very suspicious that the wording was "significantly more" instead of "statistically significant", especially when she specifically called out statistical significance for the other groups right before it. If the difference was not statistically significant, that fact is intentionally and malevolently obfuscated.

jacobolus 10y ago
This is not an ambiguous phrasing; this is a standard way of writing the summary in prose. There is nothing remotely “suspicious” or “malevolent” here. It clearly expresses that the only one of four categories whose callback rate rises to the level of statistical significance (presumably something like p < 0.05 of null hypothesis that all rates are the same) was the rate for high-class men. The reason the sentence doesn’t include the word “statistically” in the second half is that it would sound incredibly awkward, and is not necessary to repeat.
tgb 10y ago
This isn't an academic journal. But, of course, you can go read their actually published piece. What they say there is: "The higher-class male applicant had a callback rate of 16.25 percent, more than four times as high as the average callback rate for the other three applicants, who collectively generated just nine interview invitations from 235 applications, a callback rate of 3.83 percent. This fourfold difference is significant not only statistically (p < .001) but also substantively, and its magnitude is especially striking when considering the fact that applicants’ entire law school records and all academic and professional experiences were identical". So no need for conspiracy theories.
- FT_intern 10y ago
  There's no need for snark. My point still stands. Statistical significance for upper class men vs. upper class women is never mentioned. The statistics in this instance are cooked and very misleading. It does not prove the author's point. The major claim in this article is that upper class men did much better than upper class women: > Why did the higher-class man do so much better than the higher-class woman? Yet the statistic only compares upper class men to everyone else. Why did the authors choose to only mention the p-value for one category vs. everyone else? Why not other data slices? The authors clearly calculated the p-value between individual groups (from their "not statistically significant" comment). Why did they not list these values? It's a choice between negligent data analysis or intentional omission.
  - ximeng 10y ago
    It sounds like if you slice the "everyone else" further you're only going to get more significant differences.
  - MagnumOpus 10y ago
    What is your beef? The study's N and the raw evaluation numbers are even stated in the layman's article, you can even calculate statistical significance yourself if you don't trust the author's statements. And they stated that UM is significantly different from the other three, but that LM, UF and LF are not significantly different from each other. What more do you want?
  - tgb 10y ago
    > Why did the authors choose to only mention the p-value for one category vs. everyone else? Why not other data slices? Because that's the only part that I quoted - go read the academic article yourself if you want the full analysis. Let's just do the p-value comparison of upper-class men to upper-class women for you just to settle this. The null-hypothesis is that both categories are equally interviewed. Total of 18 people interview (13 of 80 men, 3 of 79 women). If all null-hypothesis is true, the 16 interviews would be randomly distributed into the men and women categories. The chance of getting at least 13 men would then be sum_(i=13)^16 (16 choose i) * (80/159)^i * (79/159)^(16-i) which equals p = 0.011. So much for malevolent omission!

jacobolus 10y ago

This is not an ambiguous phrasing; this is a standard way of writing the summary in prose. There is nothing remotely “suspicious” or “malevolent” here. It clearly expresses that the only one of four categories whose callback rate rises to the level of statistical significance (presumably something like p < 0.05 of null hypothesis that all rates are the same) was the rate for high-class men. The reason the sentence doesn’t include the word “statistically” in the second half is that it would sound incredibly awkward, and is not necessary to repeat.

tgb 10y ago

This isn't an academic journal. But, of course, you can go read their actually published piece. What they say there is: "The higher-class male applicant had a callback rate of 16.25 percent, more than four times as high as the average callback rate for the other three applicants, who collectively generated just nine interview invitations from 235 applications, a callback rate of 3.83 percent. This fourfold difference is significant not only statistically (p < .001) but also substantively, and its magnitude is especially striking when considering the fact that applicants’ entire law school records and all academic and professional experiences were identical". So no need for conspiracy theories.

FT_intern 10y ago
There's no need for snark. My point still stands. Statistical significance for upper class men vs. upper class women is never mentioned. The statistics in this instance are cooked and very misleading. It does not prove the author's point. The major claim in this article is that upper class men did much better than upper class women: > Why did the higher-class man do so much better than the higher-class woman? Yet the statistic only compares upper class men to everyone else. Why did the authors choose to only mention the p-value for one category vs. everyone else? Why not other data slices? The authors clearly calculated the p-value between individual groups (from their "not statistically significant" comment). Why did they not list these values? It's a choice between negligent data analysis or intentional omission.
- ximeng 10y ago
  It sounds like if you slice the "everyone else" further you're only going to get more significant differences.
- MagnumOpus 10y ago
  What is your beef? The study's N and the raw evaluation numbers are even stated in the layman's article, you can even calculate statistical significance yourself if you don't trust the author's statements. And they stated that UM is significantly different from the other three, but that LM, UF and LF are not significantly different from each other. What more do you want?
- tgb 10y ago
  > Why did the authors choose to only mention the p-value for one category vs. everyone else? Why not other data slices? Because that's the only part that I quoted - go read the academic article yourself if you want the full analysis. Let's just do the p-value comparison of upper-class men to upper-class women for you just to settle this. The null-hypothesis is that both categories are equally interviewed. Total of 18 people interview (13 of 80 men, 3 of 79 women). If all null-hypothesis is true, the 16 interviews would be randomly distributed into the men and women categories. The chance of getting at least 13 men would then be sum_(i=13)^16 (16 choose i) * (80/159)^i * (79/159)^(16-i) which equals p = 0.011. So much for malevolent omission!