I CANT INTO SCIENCE, I HAVE THE DUMB.

Integralmath, our Justicar, recently reposted my 'debate' with Steve Kern on his YouTube channel, and its gotten some fun comments. One was from someone making fun of Kern:

I CAN'T INTO SCIENCE, I HAVE THE DUMB.

I love it and literally loled*.

While the commentor was making fun of Kern, I also had to laugh because even though I am a scientist, I TOTALLY feel like this sometimes.

For instance, last Friday I was really busy. Experiment after experiment after experiment.

Not one thing worked.

Not one goddamn thing.

New stuff I was trying didnt work. Old stuff I have done a million times didnt work. And the best: When I was trying a new thing and I didnt think about a control I needed so I have no friggen idea if it worked (but it probably didnt). On top of it all, Bossman and I are editing a paper, and ironing out details is immensely frustrating-- Data 1 and Data 2 obviously overlap. OBVIOUSLY. But what is the right stat metric to 'prove' that they are 'statistically similar'? I dunno why 'WTF? THEYRE THE SAME DAMN THING!' isnt acceptable terminology for your results section. I get it. And then there is "Is this the right way to say that? Or will saying it that way hit someones hot-button?" "Do we even want to bring this up? Or is it just making the paper too long?" (section I spent over 9000 hours writing).

There is one phrase that sums up how I felt that day:

I CANT INTO SCIENCE, I HAVE THE DUMB.

And when you feel like this, there is nothing you can do but go home, go on a run/drink a beer/get a good night of sleep, and get up the next day and try again.

The only way you can do that, the only way you can get through the "I CANT INTO SCIENCE, I HAVE THE DUMB." moments, is if you love what you do.

So I recently gave a high school student wanting to know about being a virologist the following advice:

Modern science is all inter-connected. You can have cell biology, immunology, biochemistry, genetics, physics, mathematics, and computer science PhDs all working on the same project in virology. Do not force yourself to take microbiology courses if you end up hating microbiology and loving biochem (I never took a micro course until grad school-- micro was a 200-level easy A for premeds at my school, lol!). Do not avoid taking ecology classes because you think you have to stay in 'medical' classes (I learned 50% of what I know about evolution/population dynamics for HIV from my ecology courses).

You can pretty much do whatever you want and have a career in virology, so dont feel like you have to box yourself in to one thing or another :)

Science isnt a medical school list of prerequisites. You can do pretty much whatever the hell you love, and have a career is pretty much whatever the hell you want to study. Do what you love in science, and you will be fine**. Do what you think you are Supposed To Do To Be A Scientist, and you will be miserable.

You have to have something that makes you keep thinking about experiments after a bad day and YOU DO NOT WANT TO THINK ABOUT EXPERIMENTS. Something to get you out of bed in the morning to try everything again, when nothing went right before.

Monday: Everything was coming up Millhouse. Beautiful data. Set up all my experiments right. Everything is going to be fine. I CAN INTO SCIENCE!

* Also the one responding to someone saying I should be more like Hitchens "I dont see why she should be drunk". lol!

** You wont be rich, but you will be fine.

Tags
Categories

More like this

Thank you, this is EXACTLY what I needed to read this morning as I head in to an interview for a grad school program I'm not sure I love, but feel like I have to do. I can go in eyes open now and decide if it is for me.

Thank you!

I thought the "I HAVE THE DUMB" would go away at some point during grad school, or at least get less frequent. Here I am in the 4th year of my PhD and it's every goddamn day (except the days I don't come into lab... I like those days).

A good trick when you don't know what to right in the section of a paper is to read another paper and to copy the style. ;)

By kamizushi (not verified) on 08 May 2012 #permalink

"You can do pretty much whatever the hell you love, and have a career is pretty much whatever the hell you want to study."

That's a pretty retarded statement. Despite every body babbling on about STEM jobs, there are whole scientific professions that have been wiped out in the US (with the total support of our government). I guess "Bossman" didn't tell you that. You would be surprised how little research is being done on HIV these days in the Pharma industry. Sure there's some vaccine work going on but not too much else. I have personally watched hundreds of scientists being thrown out on the street and they are NOT working as scientists any longer. These are my friends and colleagues. THEY COULD NOT FIND A JOB IN THEIR PROFESSION! There aren't any jobs for them in the US (and they don't want to move to China - it's even more backward ass than Oklahoma.) I am glad that you are a (very) young enthusiastic researcher. I was like you once upon a time. I had a job and a career and I loved doing research. Then reality took over. I'm 49 years old and my little company is running out of money. I'm going to lose my job soon and there is no way I'm going to get another job as a medicinal chemist. I know that. Everyone I know is trying to get out of the profession because it's circling the drain. Don't encourage young people to go into science. There's no jobs and there's only so many of those academic positions that open up every year.

I'm not a pessimist. I'm an optimist with experience.

Thanks for the debating.

I liked the "if evolution was false, ..." question. I'd think we'd still be left with the body of evidence that we've observed with evolution and any replacement theory would have to explain what evolution explains. Similar to how although Newtonian physics is wrong in light of relativistic physics, but relativity has to encompass Newtonian observations and theory. If some sort of deity-explanation is supposed to replace evolution, it additionally has to explain why evolutionary theory works so well.

That's a pretty retarded statement.

"Should I be a virologist?"

and

"What do I have to do if I want to be a virologist?"

... are different questions.

I hadn't listened all the way to the end. Ha! "There is no such thing as cats turning into dogs." and "hasn't begun to replicate itself any farther than what you are describing". Kern completely misses the "endogenous" part of ERV. Having the same kind of mistakes in the allegedly different biblical "kinds" is evidence that they are of the same kind. The tree is connected, they are not wholly new and separate macro-creations.

Maybe that's the way to think of it: shared ERVs are strong evidence against the theory of "macrocreationism". To encompass that fact, the alleged deity must be a trickster who purposely blended the macrocreated "kinds" at a microcreation level to make the biblical interpretation look like a lie.

Happens to us all. Then we sit around feeling like failure frauds, Imposter Syndrome and the like. Some days nothing works. But then, later, you get a good day.

I'm sure statistics explains it all. Or a Jokester Universe, in it to make your life hard for the lulz.

Please let us know if you find a good stat metric for showing the obvious overlap of Data 1 and Data 2.

Two things came to mind:
1) Doing a convolution of Data 1 over Data 2.
2) Using some distribution divergence test out of information theory to show that both Data-1 and Data-2 have the same prior.

By bibliovore (not verified) on 08 May 2012 #permalink

Not one thing worked.

Pareto's principle definitely applies in research. Actually, in a stronger way: you get 90% of your results from 10% of your work.

Data 1 and Data 2 obviously overlap. OBVIOUSLY. But what is the right stat metric to 'prove' that they are 'statistically similar'?

What's wrong with a t-test? 'The difference between populations 1 and 2 are not statistically significant.'

By Bill Door (not verified) on 08 May 2012 #permalink

Pareto's principle definitely applies in research. Actually, in a stronger way: you get 90% of your results from 10% of your work.

So does Sturgeon's Revelation.

By D. C. Sessions (not verified) on 08 May 2012 #permalink

"What's wrong with a t-test? 'The difference between populations 1 and 2 are not statistically significant.'"

That would simply show that you are unable to reject the null hypothesis that they (well, just the means for a t-test) are the same. This could be because they are in fact the same, or, importantly, because you just don't have enough data to tell them apart.

Knowing when to say "I quit *for today* and I am going home" is one of the most essential lessons for anyone to learn. For me, I built on my lab manager's axiom: after you break your second piece of glassware you must go home and do nothing important for the rest of the day. The more things break, the more upset you are, the more things you will break.

(I will grant that this is not always possible, if your samples will be ruined by tomorrow, but you can at least ask for some common-sense supervision to make sure you don't throw out the important fraction.)

It may sound trite, but tomorrow is a new day (if you get to sleep) and a new start.

By JustaTech (not verified) on 08 May 2012 #permalink

Snap - I've been mentioned by both my first and last name: Integralmath Justicar. lol

How to prove that data set 1 and data set 2 say the same thing? Easy. For a detailed guide, please see Judy Mikovits take-home* notebooks.

My favorite part of the Q&A (that wasn't the 'debate' debate part. Suppose we should call it the Q*oA - question and occasionally answer) was when Kern was asked if Genesis could be falsified. As his temper flared, he finally said the second (out of 2) statements that was (inadvertently) exactly correct: Can it (Genesis that is) be falsified? It's being falsified all the time!

The other thing he said exactly right is that his theology determines his science. He didn't need to say it, but I appreciate the disclaimer nevertheless. =^_^=

*careless grad student/lab assistant not included.

"Do what you love in science, and you will be fine (you won't be rich, but you will be fine)." That's the story I tell to high school students or undergraduates wondering whether they should follow a career in science.

I grew up reading "old-school" SF, and I really wanted to be a "scientist." Some time during my second year of grad school, plugging away at some track fitting code on my VAXstation (yeah, yeah, I *am* that old), it hit me. "Hey! I'm a _scientist_! I'm doing what I always wanted to do."

I've never forgotten that feeling, and it never fails to re-energize me after too many interminable phone meetings with colleagues, or the seventeenth revision of a paper adding another decimal place to the B -> pi l nu branching fraction. I'm a scientist because I _want_to_be_, not because I "have to be."

By Michael Kelsey (not verified) on 08 May 2012 #permalink

"Data 1 and Data 2 obviously overlap. OBVIOUSLY. But what is the right stat metric to 'prove' that they are 'statistically similar'?"
Ehm.. Pareidolia and datamining just told me you'd best leave it out

By rijkswaanvijand (not verified) on 08 May 2012 #permalink

That would simply show that you are unable to reject the null hypothesis that they (well, just the means for a t-test) are the same. This could be because they are in fact the same, or, importantly, because you just don't have enough data to tell them apart.

That problem exists with any test, but there may be other reasons why a t-test is not appropriate, for example if the data are not normally distributed. Not sure what sort of data Abbie needs to compare, but how about a KolmogorovâSmirnov test for two populations?

Well, I have a shitload of data from experiments that no one has done before, so its a matter of 'Is this the best way to present this kind of data?' 'Is this the best statistical analysis of this kind of data?'

But the frustrating one is I have to compare two Sigmoidal curves-- I cant assume anything, like with dose-response. I cant assume 0-100%. When they are plotted on the same graph they are on top of one another... but how do you get the stats to back it up? The answer might be simple, but I have no friggen clue.

Also, everything is coming up Millhouse again yesterday and today so far-- reminded me that science is like heroin. You keep going from your lowest low, just because you need another hit. And man, when you get that hit-- that beats any fuckin cock in the world.

At the risk of sounding retarded and being on the wrong track. A chi squared test of the data and theoretical Sigmoidal curve? Something like dataset 1 chi-squared, dataset 2 chi-squared, and combined dataset 1 & 2 chi-squared.
Maybe thinking about a chi-squared test will point you in the direction of something that will help.

science is like heroin. You keep going from your lowest low, just because you need another hit. And man, when you get that hit-- that beats any fuckin cock in the world.

This needs to be the new tag line of your blog.

#17, windy
Thanks for mentioning that. The 'you need more data to see the significance' bogeyman never really goes away, neither does his life partner, the 'you're only seeing significance because of some unknown systematic error' bogeyman.

By Bill Door (not verified) on 09 May 2012 #permalink

Also, everything is coming up Millhouse again yesterday and today so far-- reminded me that science is like heroin. You keep going from your lowest low, just because you need another hit. And man, when you get that hit-- that beats any fuckin cock in the world.

I'm stealing this and taking credit for it. Just like Mitt Romney.