La Banda Borracha
January 30, 2017 11:40 PM   Subscribe

Could you help me come up with some workarounds for this computing/statistics/work problem? Not a human relations question, I promise.

My work colleague and I are hoping to submit a poster to a scientific conference. She's the brains, I'm the writer -- and a contract lab did the experimental design and carried out the experiments. However, we just realized (well, Brainy realized) we need to run our own statistics on the data, because the contract lab did something janky with a t test, and this data needs an ANOVA. So just do it, right? Well, we need statistical software. Brainy does not have admin rights on her computer, but I do on mine. I installed a trial of Statistica on my own computer. I don't know how to use the software (but I can learn!) and I do not know about ANOVAs (harder to learn quickly, I'm thinking). Brainy knows all that and more.

But Brainy is in Buenos Aires and I am in Los Angeles. And our deadline is in 8 days.

That is not enough time to navigate our company's purchasing process, and even getting IT support for Brainy to install the Statistica free trial (assuming they'd be willing to do that) would take several days. So what can we do?
  1. I teach myself the software and stats and run the analysis. Brainy would be too polite to say she didn't trust it but would she? Is this feasible?
  2. Brainy teaches me to run the analysis on my computer by web conference. Is this feasible? Or am I asking her to essentially take me to grad school?
  3. Brainy runs the analysis herself on my computer -- using that software computer techs use where they control your mouse remotely? What is that called?
  4. Try to get our big boss (who is aware of the issue) to get IT to move quickly to just install the trial software on Brainy's computer.
  5. Your own, better, idea???
Thanks for any help you can give me!
posted by pH Indicating Socks to Computers & Internet (15 answers total) 1 user marked this as a favorite
 
R and whatever remote desktop solution your company has.

Alt drop a few bucks out of pocket for an amazon workspace. I wouldn't, tho.
posted by PMdixon at 11:47 PM on January 30, 2017 [1 favorite]


You don't need a very powerful computer to run stats software and do an ANOVA. (In fact, I'd be surprised if there wasn't some sort of in-browser solution.) Does Brainy really not have any sort of private computer at home s/he can use? If so, install R on it or something and s/he can do it at home.
posted by lollusc at 11:50 PM on January 30, 2017 [1 favorite]


In fact, you can do Anovas in excel apparently. (I just Googled it, but I'm on my phone and can't link easily.) I bet Brainy already has access to excel.
posted by lollusc at 11:52 PM on January 30, 2017 [1 favorite]


1. Brainy goes to her boss to light a fire under IT's ass to make it possible to install any of a dozen free and open source applications and software libraries that can do an ANOVA

2. Brainy gets access to another coworkers, colleagues computer at Brainy's institution to do the ANOVA

3. You find a colleague at your institution who can run an ANOVA on the data.
posted by zippy at 1:11 AM on January 31, 2017


GraphPad Prism has a 30 day free trial and is very user-friendly!
posted by emd3737 at 3:08 AM on January 31, 2017


In fact, you can do Anovas in excel apparently.

This is true, but maybe not a good idea if neither of you is familiar with the method and you are short on time - particularly if you have a lot of data. Excel is notorious for letting you shoot yourself in the foot in non-obvious ways when it gets to scientific computing.

Don't you have any colleagues with proficiency and the right software that you can ask for help? You'd give them a co-author spot or a mention in the acknowledgements.
posted by Dr Dracator at 4:21 AM on January 31, 2017


Several thoughts.

I am in favor of Brainy finding a way to install software that can do an ANOVA, or finding a colleague that understands ANOVA well and can do it for you. If Brainy can't install anything but is comfortable in R, they can run R directly in the browser at http://www.r-fiddle.org/.

But assuming none of that is possible...

Various programs with free trials or web applets (GraphPad, Statistica, etc.) or software you likely have access to (e.g., Excel) should be able to do an ANOVA for you without causing too much confusion.

However.

These can lead you astray very quickly if your data do not meet all the assumptions those programs are making. Do you have the same number of data points in each group for the ANOVA? Did the structure of the experimental design introduce any subgroups into the groups you want to compare with the ANOVA? (For example, in my world, I might be exposing plants in the greenhouse to three different levels of light, but i have three different tables for high-light, three more tables for medium-light, etc - those tables are subgroups). Were the subjects paired, or were there any repeated measures of the same subject (e.g., before-and-after measurements, or measurements of the same individual over time)? I could go on, but the point I mean to make is that the way you set up an ANOVA depends on the nature of your data and the experimental design, and software can give you a very wrong answer if it is assuming you have one sort of setup when you really had another. Dr Dracator is right on in saying that you can shoot yourself in the foot with Excel (and I would argue other statistical software) if you're not familiar with the details of the method.

That said - if you just have a number of groups to compare, same number of samples in each, and no complications like subgroups or repeated measurements over time, then Excel or one of the free trials of other statistics software will probably do you just fine for the purposes of the conference. (I'd encourage you to revisit the analysis more carefully afterward though, just in case!)
posted by pemberkins at 4:28 AM on January 31, 2017


Thanks for the answers so far!
The issue is not that the full version of Statistica costs money, the issue is that Brainy cannot install anything on her work computer without IT involvement, and we have little time to wait for them to help.
Statistica is what Brainy is familiar with and says she needs, and she knows everything about statistics, so I guess Excel is out for reasons.
I love the idea of Brainy installing the free tiral of Statistica on her home computer, and I will suggest that!
It is a big company; no doubt someone has statistical software and could do the data analysis for us, but are they willing to drop everything and stay up all night to make our deadline, as Brainy and I fully expect to do?
posted by pH Indicating Socks at 4:31 AM on January 31, 2017


Brainy just woke up and emailed me -- her IT guy doesn't like that the free trial of Statistica is free, because ours is confidential data.
But a purchase order at our company takes between a few days and a few weeks! Ugh!
posted by pH Indicating Socks at 5:03 AM on January 31, 2017


How much data are you dealing with? How many rows, or how many meg?

I'd create a free trial account with one of the cloud companies, spin up an instance there, and install the free trial of Statistica on it, as PMDixon suggested. Might cost $10usd, depending on how much data you have to upload, but it'll probably be free within the monthly free limits.
posted by at at 5:42 AM on January 31, 2017


Brainy runs the analysis herself on my computer -- using that software computer techs use where they control your mouse remotely? What is that called?

As PMDixon mentioned above, it's called Remote Desktop software, and Chrome makes a clean and simple version. She may have to install a Chrome extension to use it on her end, but if that is a problem on her work computer, she could just do it from her home computer.
posted by Rock Steady at 5:57 AM on January 31, 2017


R is still the answer, you can learn how to do a simple anova on your computer in an afternoon, especially if she (or anyone) can coach you through it. I'm sure some of us could help here, and there are also myriad R discussion boards and tutorials online.

Heck I'd offer to do it myself but a) sounds like you'd be unable to share and b) my computer is out of service this week.

Good luck!
posted by SaltySalticid at 6:16 AM on January 31, 2017


Just a thought...is there any chance you're comparing two groups? Since an ANOVA and a t-test are equivalent for two-group comparisons, there's a chance the contract lab's results would be okay, despite having used a t-test.
posted by MrBobinski at 5:21 PM on January 31, 2017


If that's not the case, here's another vote for R. I like Quick-R a lot, which has an ANOVA walk-through.
posted by MrBobinski at 5:23 PM on January 31, 2017


Just to let you know how it turned out...
The simplest solution was best: Brainy got the trial software on her home computer, did the work, and found out that our conclusions in the poster, which had been based on the flawed stats, actually held true for most of the data when the stats were done properly! Enough to make a credible poster, anyway, and none of this "a trend was observed that did not rise to the level of significance" crap either! Yay!
Thanks for your help!
posted by pH Indicating Socks at 6:23 PM on February 2, 2017 [1 favorite]


« Older Do we need to get married before things get worse?   |   What to read after Frog and Toad? Newer »
This thread is closed to new comments.