Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

One in five genetics papers contains errors thanks to Excel (2016)

science.org

Post reply on HN
Print view View original

121–130 of 267 posts

Re: One in five genetics papers contains errors thanks to Excel (2016)

#121

Post by eviks »

bikenaga wrote: post #70
cycomanic wrote: post #65

Earlier quoted context omitted.

I had a discussion on one of the Nim boards because Nim does the same, i.e. Writing -2^2=4 but 0-2^2=-4 because Nim treats the first as the unary - which takes precedence over exponentiation. I realise that you can argue how it is technically correct (and interestingly several of the people I was talking with couldn't even get my point), but I still argue it's incredibly unintuitive for anyone coming from an regular…

You're right that violating established mathematical convention can be a deal-killer for some kinds of adoption. A few years ago I was reading the docs for a new programming language, thinking it might be useful in teaching. The docs were well-written and in a beautifully produced book. I got to the chapter on trig functions and discovered that they'd decided to make angles increase clockwise . And there was a graph…

"more natural" is a good reason any time since otherwise you get to perpetuate these bad conventions for all the future generations to suffer.

Re: One in five genetics papers contains errors thanks to Excel (2016)

#122

Post by eviks »

FabHK wrote: post #103
bikenaga wrote: post #48

An annoyance with spreadsheets that deterred me from ever using them in teaching is that they've perpetuated a arithmetic order-of-precedence bug. ("Bug" in the sense that it contradicts long-standing mathematical convention.) If you type -3^2 in a cell and press ENTER, the spreadsheet tell you it's "9". It should be "-9"; in math, exponentiation has precedence over unary minus, so you square 3, then negate the resul…

I recall reading that this was easier to implement in memory constrained machines running Excel back then, and now can't be changed for compatibility. It has bitten me when I computed the pdf of a standard normal in Excel, invoking exp(-A1^2), say. Someone made a website (in 2003, it's a bit out of date) tracking this issue: http://www.macnauchtan.com/pub/precedence.html

But of course it can be changed, you can make it configurable even if you decide the default should continue be this silly mistake "for compatibility"

Re: One in five genetics papers contains errors thanks to Excel (2016)

#123

Post by eviks »

theodpHN wrote: post #44

So, the question is: Would there be fewer or more errors if something other than Excel was used? Has much rigorous research been done on this?

Unfortunately researchers can't fork the universe with a different set of defaults for this rigorous study you appear

Re: One in five genetics papers contains errors thanks to Excel (2016)

#125

Post by eviks »

lukan wrote: post #37
hilbert42 wrote: post #35

Don't we ever learn. We've known since VisiCalc in the '80s not to trust spreadsheets for important data without crosschecking the results. The question is why do we still use substandard tools for processing important data like this.

"The question is why do we still use substandard tools for processing important data like this." Because there is no better alternative (yet)? A better alternative needs to be really better, to justify the effort of people relearning how to do things in this better tool then.

There isn't much relearning needed since errors like the one described in the article illustrate that people don't know much about this tool either

And Sheets isn't "really better", yet gained a noticeable share

Re: One in five genetics papers contains errors thanks to Excel (2016)

#126

Post by remus »

jimnotgym wrote: post #124

Fixing the headline... One in five genetics papers have errors caused by mistakes in the use of Excel As if longhand calculations never have errors?

The difference here is that the excel errors are systematic so should be more easily avoidable, whereas long hand calcs are more varied and harder to correct.

Re: One in five genetics papers contains errors thanks to Excel (2016)

#127

Post by verbify »

eviks wrote: post #122
FabHK wrote: post #103

Earlier quoted context omitted.

I recall reading that this was easier to implement in memory constrained machines running Excel back then, and now can't be changed for compatibility. It has bitten me when I computed the pdf of a standard normal in Excel, invoking exp(-A1^2), say. Someone made a website (in 2003, it's a bit out of date) tracking this issue: http://www.macnauchtan.com/pub/precedence.html

But of course it can be changed, you can make it configurable even if you decide the default should continue be this silly mistake "for compatibility"

It is configurable - with brackets.

Re: One in five genetics papers contains errors thanks to Excel (2016)

#128

Post by eviks »

rectang wrote: post #11
golergka wrote: post #4

I studied bioinformatics in university back in 2005, and the first subject we were trained on was not Python or Java, but using Excel. Our teachers explicitly told us about excel's conversions to dates and other pitfalls. I find it mind-boggling that some researchers still made these mistakes 10 years later. Excel is a wonderful tool. But you need to learn your tools and find out about possible footguns.

Haranguing your users not to make mistakes is utterly worthless. The only way to move the needle is to design systems that are easier to use correctly.

Do you think the comment is written by Microsoft??? The sloppy researchers are not his users. Another way to move the needle is to place higher value on technical competence.

Re: One in five genetics papers contains errors thanks to Excel (2016)

#129

Post by eviks »

verbify wrote: post #127
eviks wrote: post #122

Earlier quoted context omitted.

But of course it can be changed, you can make it configurable even if you decide the default should continue be this silly mistake "for compatibility"

It is configurable - with brackets.

You confuse configuration with data entry

Re: One in five genetics papers contains errors thanks to Excel (2016)

#130

Post by the_gorilla »

prepend wrote: post #18
rectang wrote: post #11

Earlier quoted context omitted.

Haranguing your users not to make mistakes is utterly worthless. The only way to move the needle is to design systems that are easier to use correctly.

Excel is not a genetics tool. It has millions if use cases, many more important than genetics. Excel didn’t care about this. It’s like people complaining because sugar gets misused. Or that murderers stab people with knives. The solutions isn’t to “fix” knives.

This is an awful mindset and I'm tired of people acting like tools only ever have pre-defined purpose and you need a special purpose tool to do anything. Also that's a dumb analogy. Here's a better one: A knife that breaks 20% of the time when you try to cut pork, because you didn't buy a separate pork-cutting knife.

I'm not doing anything nearly as special and always have dates import as numbers for whatever reason. Thanks microsoft.

Post reply on HN