The Secret Of Scientific Research – What They (Normally) Don’t Tell You

The height of the trees were not deemed important as it would require standing in middle of a freeway to measure.
-Ian Smith ‏@EcologIan

1. Choosing A Methodology

Statistical advice was integral to the initial experimental design, but we decided the advice led to too much work.

– David C Logan ‏@angerstusson

Selected to use open data for analysis because we didn’t manage to negotiate access to potentially more useful sources.

– Muki Haklay ‏@mhaklay

A Northern blot was run instead of realtime QPCR because the PI is old and does not trust results unless he sees a band.

– Adriana Heguy ‏@AdrianaHeguy

I used a snowballing method to gather interviewees/I just asked my mates as no one else came forward.

– Steven Vaughan ‏@lawvaughan

We used default priors because we have no idea what the fuck they mean, and fiddling with them changed all our results.

– rafael maia ‏@hylopsar

2. Carefully Logging Fieldwork

N=99 because the 100th etherised fruit fly became conscious and walked off.

– Dane ‏@danientifically

Samples from this hotspring are not represented because they were eaten by a bear as we watched in horror.

– Russell Neches ‏@ryneches

We carefully decontaminated our sampling spoons, but did nothing about all the slag dust blowing around on site.

– Maria ‏@mcmoots

We only measured it in 5 patients cos the software crashed/patient moved/I pressed the wrong button the other 20 times.

– Jude Dineley ‏@judedineley

We only collected one core per tree because why don’t you try repeatedly screwing a hand-powered drill into solid oak?

– Scott St. George ‏@scottstgeorge

3. Precision in the Lab

The samples were incubated at 37 degrees for the duration of one episode of the Daily Show.

– twoearsoneheart ‏@twoearsoneheart

DNA band was excised from the gel, dropped on floor and then scraped into an epitube for extraction.

– Kenneth Bruno ‏@KSBruno9

Protein digestion was performed using an automated system. Glitches were solved by hitting the robot with a spanner.

– Vibhuti Patel ‏@VibhutiJPatel

Samples were stored at -80 degrees until required except for a day or so when our freezer compressor failed.

– David C Logan ‏@angerstusson

The eppendorf tubes were “shaken like a polaroid picture” until that part of the song ended.

– Rebecca Weinberg ‏@sciliz

4. Evaluating Findings

We use this simple, intuitive method because we were scared of asking a real statistician how to do it properly.

– Philip Bett ‏@pbett

We chose this model because we already tested the drug in 3 others but it didn’t work.

– Hannah Mace ‏@MaceHannah

Subjects A, C, and F were excluded because they were fucking up the results we wanted.

deonandan ‏@deonandan

The appropriate statistical analysis was selected using the “which one gives the smallest error bars” method.

Henry Scowcroft ‏@oh_henry

Hypothesis was not redrafted to fit results, and results were not cherry picked to fit hypothesis. It was more mutual.

– Snarky Everything ‏@BadPhysics

5. Presenting Them Thoughtfully

Most of the buzzwords in the title have nothing to do with the article, but they wont publish it without them.

– James ‏@jholyhead

Our methodology is tentatively relevant to a range of other disease sites because saying so might get us more funding.

– Henry Scowcroft ‏@oh_henry

LITERATURE CITED (but not actually read).

– Dr. Twittenheimer ‏@DrTwittenheimer

We didn’t read half of the papers we cite because they are behind a paywall.

– Georgina ‏@geocorner

From my supervisor: bear-bear interactions were significantly reduced in late season, after one of the two bears died.

– Tom Houslay ‏@tomhouslay

Last But Not Least

What we lack in meaningful results we make up for with a witty pun in our title.

– Dr. Twittenheimer ‏@DrTwittenheimer

Hooked? Check #overlyhonestmethods for more.