Earlier this week, a friend and I were talking at lunch about some professors who uncovered research fraud in a Harvard Business School study: https://datacolada.org/109

This article is kinda fascinating, in at least two respects:

1. I learned things about metadata in Excel that I never knew*;

2. The subject of the study, while interesting, hardly seems like a hot-button issue, or a career-making one. To my apparently naive eyes, it doesn't seem like the kind of research in which the stakes would be high enough to motivate folks to fake the results. I guess, as the (great) song** says, it goes to show you never can tell.


*In the past I used to advise folks that the surest way to strip metadata from a document is to print it out on paper, and then scan the paper copy into a new file. Nowadays, maybe even that doesn't work!

**Speaking of which, the famous video of Bruce Springsteen's impromptu performance of said song at Leipzig in 2013 is one of the greatest exhibitions of live musicianship in the history of popular music: https://www.youtube.com/watch?v=L-Ds-FXGGQg

I just realized I labeled this post as "OT", and I have two questions:
- Is there such a thing as an OT post on the Quad?
- It will be interesting to see if this topic manages to escape the Quad's Iron Law of Thread Convergence. .

(01-15-2025 05:59 PM)georgewebb Wrote:  I just realized I labeled this post as "OT", and I have two questions:
- Is there such a thing as an OT post on the Quad?
- It will be interesting to see if this topic manages to escape the Quad's Iron Law of Thread Convergence. .

Luckily, there is a good chance that no president in the last decade has ever touched an Excel spreadsheet!

I started to read that post this morning, but didn’t have the stamina to get that into the weeds on Excel. But maybe another day.

That said, my gut tells me that a study that isn’t earth shattering, but that would support someone’s overall research focus and would get published, is actually the research most likely to get fudged. A high profile study would be more likely to be thoroughly scrutinized. But publishing is critical to academics, so i see an incentive to fabricate data if a study was not as compelling as one hoped. From what i did get through, it sounds like the prof in question did just that - took a data set that wasn’t compelling and adjusted it to make it compelling.

(01-16-2025 06:44 AM)RiceLad15 Wrote:  Luckily, there is a good chance that no president in the last decade has ever touched an Excel spreadsheet!

I started to read that post this morning, but didn’t have the stamina to get that into the weeds on Excel. But maybe another day.

Well, the article is certainly weedy, or weedful, or however one ought to say it. :)

(01-16-2025 06:44 AM)RiceLad15 Wrote:  That said, my gut tells me that a study that isn’t earth shattering, but that would support someone’s overall research focus and would get published, is actually the research most likely to get fudged. A high profile study would be more likely to be thoroughly scrutinized. But publishing is critical to academics, so i see an incentive to fabricate data if a study was not as compelling as one hoped. From what i did get through, it sounds like the prof in question did just that - took a data set that wasn’t compelling and adjusted it to make it compelling.

Interesting point!

I'm reminded of one of my favorite quotes... Figures don't lie.... but liars figure.

There is an art to taking data that says one thing and changing it to appear to say something else. Begin a look at the economy or inflation or all sorts of numbers on election day and it may show a dramatically different result than if you look at inauguration day, or 'initiation of first budget' day. Climate studies look very different if you start in 1970, or 1890, or 1400, or 4,000,000 BC

those are just simple adjustments not involving falsifying data.

(01-16-2025 05:04 PM)Hambone10 Wrote:  I'm reminded of one of my favorite quotes... Figures don't lie.... but liars figure.

There is an art to taking data that says one thing and changing it to appear to say something else. Begin a look at the economy or inflation or all sorts of numbers on election day and it may show a dramatically different result than if you look at inauguration day, or 'initiation of first budget' day. Climate studies look very different if you start in 1970, or 1890, or 1400, or 4,000,000 BC

those are just simple adjustments not involving falsifying data.

I had a patent case in which there was a fair bit of argument about whether a particular graph should be shown on a log-log plot or a log-linear plot. There were good arguments either way. The log-log plot obviously compressed the scale, but it also made the physical property in question easier to see, so qualitatively it was more revealing.

Which reminds me:
One pet peeve is the popular but maddening trope that "traditional" (by which is really meant Mercator) maps are "inaccurate" or "wrong". As everyone in this forum knows, Mercator maps (like all map projections) are accurate in what they are designed to be accurate in (in this case, compass directions), and certainly not across-the-board "wrong". To say that a particular projection is blanketly inaccurate is like saying abstractly that the color blue is inaccurate; it may not be the color you want for a particular purpose, but no color is inherently right or wrong.

Update: the professor in question has been canned:
https://nypost.com/2025/05/27/us-news/ha...in-studies

Her website has some detailed documents and rebuttal arguments:
https://www.francesca-v-harvard.org