Modelling vs. Experimental work

[virtua]

Which direction you chose and why? What are the differences, difficulties, advantages/disadvantages of each one?

[pealio]

I'm actually doing both. Found a great systems biology program where I have co-advisers. One is strictly theoretical modelling while the other is strictly experimental. Collaboration with both is allowing me to experimentally test/back-up my models!

However, if you want an opinion on one or the other....

Modeling has an advantage in that you can publish papers much much faster. There is no experimental work. However, this leads me to the disadvantage that you have no experimental work to back-up your claims. Therefore, a lot of your criticism from reviewers will be about it being theoretical. Also, because you have no experimental work, your papers most likely will be lower impact which isn't good.

On the experimental side, you will probably have higher impact but reviewers will be much more critical of your techniques. Also, experiments take forever and you need a lot more funding for them.

[GeoDUDE!]

I disagree that you will have higher impact in experimental vs theoretical: if you are doing truly novel work it won't matter. I also disagree thta modeling allows you to publish faster. One of my simulations takes about 3 months to run on ~500 cores. This isn't that uncommon, this is why there are several NSF funded high computing centers.

I actually don't think it matters which you pick.... though I picked pure modeling (over an offer to do both analog and computer models) because I thought while doing both would be cool, the computational skills I am learning now give me a better back up plan : I can work for companies like google, goldman sachs or even bloomberg. 

[lostongilligansisle]

I think it depends more on personal preferences - I'm an experimentalist at heart, and always have been. I would not be happy doing modelling. (Although I can't argue with geoDUDE! about having some excellent skills that can be applied elsewhere.) Either way, I think pealio's situation is the best way to go: get involved in a collaboration that employs both - models can be used to narrow the experimental resources required for the most likely desired outcomes,  and experimental results can be used to build better models. It doesn't mean you have to do both -- that would likely be too much for one grad student, depending on the project.

[Monochrome Spring]

The majority of my research is experimental, and I chose that because I like to manipulate what is going on my studies and work in very controlled greenhouse settings. However, I also have included a modeling component to my dissertation, because I want to be well-rounded as a scientist. My advisor has done a mix of the two in the past, although his current research focuses mostly on experimental.