Totally bogus, the survey was an online poll conducted by the very website that was reporting it, who has lied about other things as pointed out by the article itself and these comments.
>Mark Kantrowitz, a well-known student loan expert, said he suspected there might be a connection between LendEDU and Student Loan Report because both claimed a statistical degree of precision from their surveys he said was inconsistent with their sample sizes.... "It stood out as something both sites do, but no one else does," Kantrowitz said. "The use of two decimal points of precision also suggested that some of their surveys were complete fabrications."
I'm not a statistics expert, so could someone tellmy why using two decimal points of precision suggests survey data fabrication.
> I'm not a statistics expert, so could someone tell my why using two decimal points of precision suggests survey data fabrication.
If you use too many decimal in a report in the physics lab the T.A. will yell at you and ask you to redo it.
I'm not sure that the two decimals is a sign of fabrication, but it's a sign of sloppy work, like someone that just copy all the decimals that Excel shows in the screen, probably because they don't know enough statistic or they don't care, and I'd take all the conclusions with suspicion, and even take a look at the methodology to collect the data (and try to ensure the data is not fabricated).
To get a result with two decimal points, let's say something like 26.21%, and be sure that the 1 has some meaning and it is not a just a blob of ink you need a lot of subjects in your sample. A naïve analysis is to have at least 10000. If you have only 103 you get the last 1 in the division but it's meaningless, you can't expect that if you make a bigger and better pool you will get a result that has a 1 there. But it is worse, because the confidence interval decrease like the square root, so you need something like 10000^2. (There are some constants here and there, so perhaps you can get away with a smaller sample, like 10000000 instead of 100000000, but I'm never remember the details, but 100 is not enough.)
https://www.chronicle.com/article/No-Students-Probably-Aren-...
Also this
https://www.cnbc.com/2018/04/25/lendedu-ceo-says-survey-spok...
>Mark Kantrowitz, a well-known student loan expert, said he suspected there might be a connection between LendEDU and Student Loan Report because both claimed a statistical degree of precision from their surveys he said was inconsistent with their sample sizes.... "It stood out as something both sites do, but no one else does," Kantrowitz said. "The use of two decimal points of precision also suggested that some of their surveys were complete fabrications."
I'm not a statistics expert, so could someone tellmy why using two decimal points of precision suggests survey data fabrication.