I wish you luck, but I long ago declared defeat against Excel. I can't prevent everyone from using it. I have come to accept that whatever Excel does is what I have to live with, because someone, somewhere out there hasn't heard the anti-Excel gospel yet.
Totally relate to this.
I studied a degree because I wanted to rid the world of Excel.
Just like I wanted to rid the world of pie charts.
Turns out the rest of the world is quite happy using both of them, no matter whether I think it's right or not.
If you produce, touch or directly act on data, Excel is no longer allowed. On the other levels the people that still use it (because they think they have the skill - which they often don't) can't do meaningful direct-action on those files anyway so that solves part of the problem of acting on bad data or bad data processing.
We do allow web-based spreadsheets, that has the benefit of limiting what you can do and drives people to either put a request in with the correct team or take internal courses on how to do it right.
Some people hate it, but then again, some people hate Excel. The difference is that the people that don't hate it now have an empowered skillset that they didn't have before. In theory you could do the same with Excel-oriented internal courses, but we didn't want to do Excel because a lot of us think it's stupid (often simply for emotional reasons but the push was based on hard data like process metrics and user input).
To add to this: it's probably because I find myself attracted to data-driven organisations where using Excel didn't help the workforce much anyway, and the positions I'm in allow for a certain degree of influence on the tools that we support and how much explicit allow/deny we apply on them.
Not knowing how to use the primary tool in your field because you only know how to use a web-based spreadsheet with 1/100th the functionality is not an "empowered skillset." It's a cruel trick to play on employees, who will have difficulty finding further jobs in that field because they don't have a basic understanding of the tool everyone else uses.
It's not a cruel trick and even if it was: web-based spreadsheets might not have the same functionality, but the people that were using the desktop-based spreadsheets weren't doing much more than basic computation on fields and search/replace. That works fine on the web.
It's not like we took away their toys and gave them dirt instead. We just made everyone take the startup course to get them acquainted and if that went well we would decide our next steps. It went well, and the next steps became removing Excel from the standard workflows.
There is tremendous competition in the spreadsheet business, and anyone is free to use Julia, R, whatever they like.
The only reason they want to use a tool that doesn't really fit their use case, is because they're not aware of better tools. I find it very hard to believe that scientists actually want to use a tool that corrupts their data. If that is true, then that is absolutely a problem with their attitude towards science and data.
It's not that hard to imagine a tool that can do exactly what Excel can, but without corrupting your data. It might even exist already. LibreOffice got mentioned a lot; it can do almost(?) everything Excel can, but without corrupting your data. If there are problems with it that make it useless to scientists, there's a good chance the LibreOffice community can fix them.
I think better tools that preserve the integrity of their data are absolutely the solution here.
Also, I don't think it's a very minor issue if the tool you use to process your data, changes your data. It may only do that in limited cases, but the fact that it does it at all should alarm anyone who cares about data integrity.
This is what I react to. Sorry I missed that you're not the original commenter, that's my bad.
I am not sure though what other tool should the biologists learn - they usually know R and Python and they use Excel because they need what Excel offers; there isn't any other software that allows them to easily do whatever they need to do with the data, quickly iterating on the ideas AND then present it nicely.
The solution here should be to fix Excel, but it really is a small issue (that has been worked around, too; and if you know how to use excel correctly, it does not happen to you) compared to making another Excel, which seems like a monstrous task.
You're claiming that they do know the alternatives but the alternatives fail in worse ways than Excel does.
I don't know what the magic sauce is that makes Excel so much better than the alternatives, or why it would be such a monstrous task to replicate that. My impression is just that to many people, Excel is simply the default tool to enter data in, no matter what the actual problems with Excel are. Excel is like gravity; good or bad, you put up with it because it is what it is. And if that's the case, that'd be pretty bad and in need of change, because Excel is not at all like gravity; it's merely one tool of many. If there's a more appropriate tool, people should use that. If there isn't, one could be developed.
I stand by my point that a standard data collection tool that modifies the data you put into it is a really bad idea for any field where accuracy and data integrity is important, and I'd expect science to be one of those fields. You may be able to work around the limitations of the tool, but the risk is still there.
123456789876, 987654321234
After opening in Excel it will convert these to:
1.23E+11, 9.87E+11
Then resaving the file (as a csv) will store the values in scientific notation and lose all of the digits.
number date currency
A B C
I'm newly returned to (very minor) Excel use and assumed formats with no feedback seems wrong. My situation is dates _not_ being picked up as dates and it not being obvious that it hasn't; I'm using ISO format (eg 2020-06-17), I guess you have to create a new date format or use USA-ordering before it will recognise dates.
At the moment users need to remember every time they import the CSV to specify A data type for each incorrectly handled column. Suffice to say they don’t always remember... and with large CSVs with hundreds of columns it’s just a pain.
This would allow existing CSV tools that work with data files to continue to work without modification, while CSV Schema aware tools could make use of the schema.
Excel should know any number that is longer than it can internally support should be stored as text. Or they need to introduce a data type that can store arbitrary precision numbers.
The asinine thing happens when you save it back as a csv. Excel doesn't think to itself "hey, I should write these full values I have stored in each cell to the csv file". It just writes whatever is currently displayed to the csv file instead.
you are 100% correct. and yet it will continue to happen.
Intellectually, it feels snobbish to single out Excel like this. I'm a software engineer in science, and I generally agree that scientists should learn some kind of coding. But you can make mistakes in Python or R as well - not to mention in physical experiments. We should check data and methods in general, not presume incompetence if people use one tool.
It's not coincidental that big evil FDA/pharma requires people to sign - on paper - their lab notes (and to have lab notes, and to sign that they calibrated the equipment, and that they checked for common mistakes).
And yes, I know this costs a lot, and that this is sort of a pipe dream. And I'm not saying this is the best solution. Renaming genes might be a lot better, because it just allows the field to move on, and Excel will eventually die out, or at least researchers will move away from it - maybe to Jupyter, if we're lucky.
So, all in all, of course Excel is just a symptom, but society is already pretty stubborn when it comes to financing R&D and progress.
Ideally excel would change, but since we know it won't, and we want to work with lots of people with minimal problems, we must adapt.
Researchers should use other, more appropriate tools, or at the very least specify the column type when importing data into Excel. It's not that hard.
It's one thing to change typed in user text in real time. That's not causing any problems. It's another to randomly mutate cells amongst tens of thousands of rows. I don't think that has ever helped anyone.
If I can get a trained immunologist looking at my data, I'd much rather have 5 more minutes of their analytical skills than teaching them about common data exchange pitfalls.
https://en.wikipedia.org/wiki/Locus_(genetics)
around the late nineties the obsession with videogames and netspeak took hold, and the popular names began to crop up.
If someone types MARCH1, DEC1, OCT4, SEPT1 into a column, why should Excel not treat these as dates? Probably 99.999% of users would prefer this auto-understanding. To make users have to then select cell formatting and select date is a terrible UI decision.
It's odd to call such a useful feature a bug.
>Excel is fundamentally not suited to analysing *omics data
But it's excellent for a massive range of basic analysis and especially for a unified platform to share results, which is why everyone uses it. It's good for publication or proposal steps of creating graphs from data output from other more specialized systems, that usually lack the flexibility and widespread use that Excel has.
If my group uses some specialized, often in house, analysis software, my results cannot be spread to others unless I put them into a more common and accessible format. And nothing in the genomics and wider scientific community has anywhere near the widespread availability as Excel.
You might as well complain they also use word docs instead of Latex or troff. Not everyone wants to spend days fighting software instead of doing work.
It should treat these as dates, unless told otherwise. But Excel goes one step further, and throws away the text you originally typed in. That means, if you type "DEC1" into a spreadsheet, see it looks stupid, and change the data type of that cell to "Text", then it's too late, and Excel will show "37226". It has destroyed the original data and replaced it with the numeric equivalent of the date.
The data type of the spreadsheet cells should be a display format only. The text you typed in (or loaded from a CSV file) should be preserved, and the data type of the cell should only dictate how that is treated and displayed, not how it is stored.
However, when you input something into a cell, Excel has to parse it and decide what value and datatype to assign to it. That is just inevitable, and the same everywhere: in Python, `x = 3` assigns an integer, `x = 3.` a floating point, and `x='3'` a string. Similarly hints (or explicit specifications of type) are possible in Excel.
However, given that, what you typed originally is not recoverable anywhere (Python doesn't record whether you wrote `x=3.` or `x = 3.00`, why should it?).
So, to the extent that it is reasonable, Excel does what you request, and to the extent it doesn't do it, it's not reasonable, I'd say: A cell need not maintain the original pre-parsing input.
That’s even worse, since a user assumes the cell holds what it shows. Excel, for example, auto completes entries, which is extremely useful for data entry, and for accuracy. If it only held the few characters entered, now it has to repeat exactly the same autocompletes every time it opens, and has to do so even if the data that originally caused the auto complete changes. All this is nearly algrithmically I’m-Seville without changing the document into a history. And now CSV is useless for data transfer.
It’s vastly more sane to change to the data the user meant to enter, show it to the user for confirmation, and save that.
That the 99.99% use case bites the remaining 1 in 10,000 is a perfectly reasonable trade off, done throughout society in uncounted ways.
Automatically converting SEPT1 to a date caused far more damage than the benefit of the fleeting convenience.
The damage that this "feature" caused and will keep causing goes far beyond the domain of bioinformatics.
Even in bioinformatics things have not been "solved", for the next decade the same gene will have two names when looking it up in previously published data.
Who are you to judge? For the novice, even for most users it'll be what they want. And for the expert (or even intermediate user) it is trivial to override.
20% of publications had errors introduced by Excel in them! The real number is probably much higher (Nature had a rate of 35%). What is the rationale in saying that the benefits must still be worth it?
See: Gene name errors are widespread in the scientific literature
https://genomebiology.biomedcentral.com/articles/10.1186/s13...
The root of the problem is not about how Excel displays information, or how Excel parses the data but that the data it tacitly changed upon saving it.
Open a CSV file that has SEPT1, don't take any action, save it right away. Boom, the data has been changed and you cannot recover the original information.
It is super hard to figure out a way to import data into excel without triggering auto-conversions. I knew a way a few years ago when I was dealing with lots more biologists, but I have forgotten now. If I had to do it now, I would write XLSX directly then manually check the file in Excel, because Excel is fundamentally untrustworthy.
I respectfully disagree with this. Excel is fundamentally not suited to analysing *omics data. It's often the default program affiliated with a .csv filetype on people's computers, but trying to get an entire field of scientific research to rewrite itself based on its glorified bugs is...wrong, in my opinion.
If you see wrong things in the world, do you accept them as they are, or try -- however ineffectually -- to force change for the better? I for one bang the drum into the wind and try to get biochemists off it. I teach people to be very sceptical of excel in my stats courses, for example (aside from some showstopping bugs and downright dangerous defaults, its RNG is particularly crap).