/sci/ - Science & Math

I still don't see why so many people are butthurt about diets. Just eat whatever you're comfortable with, and whatever seems most healthy. Forcing diets onto others is stupid as shit.

I switched to a vegetarian diet about a year and a half ago, but recently switched to a bit more flexible diet (I won't refuse fish/meat if someone has cooked me a meal, or when I want to try a local specialty when abroad, for instance). Though I wouldn't mind seeing others reduce their fish/meat consumption (not too much, since we probably need animals for leather and other stuff anyway), I honestly don't give two shits about what people do with their diet, as long as they're happy with it.

Now go have your steak or block of tofu or whatever you like.

>>5396470
>Physics is neat, but be aware that a lot of it has very little to do with reality. Spherical chickens in a vacuum? Hardly a REAL science.

Stop spewing bullshit about a subject you know nothing of.

>>5396404

Picked out three books that seemed prettty simple. Enjoy!

http://www.mediafire.com/?j0qe7cw4p5w8vnw

>>5396404
Should have plenty of those. I'll what novice/intermediate books I have, and I'll upload them in a sec.

>>5396123
/sci/ always acts like a bag of dicks to any Economist, mainly because it's not an exact science (and probably never will be). Anyone who says "Economics is just History/Psychology/something else" has no clue what Economists actually try to do (i.e., 95% of /sci/)

>>5396070
pro-tip: It does follow the scientific method. Hypothesis -> Test -> Publish -> Reproduce. We simply have less certainty about results than our natural science bros.

>>5396252
Worst insult you can give an Economist is to call him a Sociologist.

>>5396156
There's no general Economics book. The field is divided into many sub-fields. If you truly want to learn something about the subject, you'll pretty much have to get a book on every topic and then figure out which topic interests you the most. I've got a 1.2GB folder full of Economics eBooks. Let me know if you'd like me to upload a few of those books.

This is an intermediate result I got while working on a project recently. Fun times.

Might be coming in late here, but whatever.

A Maximum Likelihood Estimator kind of takes the opposite approach of the OLS method (though the results can coincide under special circumstances). Rather than taking a dataset as a representation of the population as given, and trying to find what characteristics this population might have when it comes to specific variables, you start off from a distributional assumption and try to find the parameters that would make finding your dataset the most likely.

Basically, you try to find out what parameter values would make obtaining your specific sample as likely as possible.

The use of a likelihood function is necessary during this procedure. Based on your distribution (Bernoulli, Normal, Gamma, whatever you like), you formulate a function for each of your observations. This likelihood function basically gives you the probability of obtaining a specific set of observations (contingent on values of the dependent variable), and the function is dependent on a parameter vector. By maximising this function with respect to the parameter set, you find out what parameters would maximise the probability of having your specific sample.

Hope that helped a bit.

>>5173607

Yeah, Durbin-Watson exists. A test for multicollinearity doesn't, though; that would just give you an error of having a singular matrix when doing an estimation (generally when you have too many dummy variables)

Where are you studying, btw, Econ-bro?

Don't use flashcards, because I have to understand how to get at solutions rather than just memorising some definitions. I only really see the appeal of them for medicine students, since their entire field of study is based on memorising shit.


>>5172549

>Durbin-Watson
>Multicollinearity

You do know that multicollinearity implies a non-full rank matrix of explanatory variables, right? That is, the determinant of that matrix is zero. Why the fuck would anyone devise a test to see if a determinant is zero if it's just part of the ordinary calculations?

>>5171301

Typical anon. Flames a tripfriend for telling the truth. Scroll up, and notice that there's only one person who said he was taking 'mathematical statistics'. Where are your posts (plural, faggot) now, babby?

inb4 linking to undergrad folks.

Coming back to OP's post: There's a whole lot of calculus in statistics, so being good at that should help considerably with statistics, unless you're just doing Bernoulli shit. Things like proficiency in real analysis and linear algebra are indispensable.

>>5171074

You'd best be joking.

Statistics get pretty damned complicated once you start doing more advanced stuff (and by that I mean beyond babby's first undergrad stats course, which is the most anyone in this thread has taken).

As much as I'd like to try, I'm not a physicist. I don't even know what the first question is asking of me.

>>5164642

Alright, awesome. Thanks!

>>5165079
If you define 'science' as any form of satisfying curiosity, then I agree as well.

>>5164495

It's an Optimal Control problem.

>>5164512

Well, I've got a system of equations and the variables within it are interconnected, but due to one of the equations being solvable really easily and through the FOCs of my Hamiltonian, I can determine the time paths for the remaining variables one by one by substituting already-derived time paths into other equations. In the end I'm left with some (rather long) equation in the form of <span class="math">\dot{x}(t) + a(t)x(t) = b(t)[/spoiler] where x(t) is not a vector, a(t) is rather simple, but b(t) is something like <span class="math">b(t) = Ae^{at} + Be^{b-c}t + C[/spoiler] (much more complicated, of course, but similar in structure). Finding eigenvalues for the system would be a pretty massive clusterfuck, so the clumsy method I'm using is sadly a lot easier to do.

Effectively, I'm wondering whether applying the afore given formula is valid for all forms of b(t), or whether there's an alternative method for when there are exponential functions within the b(t) (i.e., I'm simply not sure how general the formula is).

Pardon if all this sounds silly or trivial; I'm simply not particularly experienced with Optimal Control problems.

Bump. Anyone?

>>5163913

Bah, obligatory LaTeX fuck-up.

It was supposed to say <span class="math">b(t) = Ae^{at} + Be^{(b-c)t} + C[/spoiler]

I'm working with a system of differential equations atm, and just want to make sure I don't do anything stupid while doing derivations. Google results weren't particularly helpful, so I figured I'd just ask here.

Suppose I have something in the form of <span class="math">\dot{x}(t) + a(t)x(t) = b(t)[/spoiler], then one can derive that <span class="math">x(t) = I(t)^{-1} \int I(t)b(t)dt[/spoiler] where <span class="math">I(t) = e^{\int a(t)dt}[/spoiler].
Is this still a valid formula to use if we have something like <span class="math">b(t) = Ae^{at} + e^{b-c)t[/spoiler] or do I have to do something different then? I don't see why it wouldn't work, but I'd rather not spend ages working things out that aren't correct anyway.

>>5136507

I can see both and can alternate between the two whenever I like. Like this illusion.

Try out Inkscape. It's a free program that's great for vector graphics.

http://inkscape.org/

>>5081852

Advanced Open Water? Don't worry then. That stuff is fine, and good fun. Enjoy it, OP! I wish I could dive again ;_;

>>5081560

Ah, quite right. Well spotted. I just typed up a wall of text to explain why I summed it, but then I noticed I misread a plus sign where there was really a minus sign.

OP, the covariance terms containing Y in them should have a minus sign in front of them instead of a plus.

>>5081527
>>5081525

Why does deleting posts no longer work? In any case, pardon the unintentional double post.

The variance of a variable <span class="math">X[/spoiler] is calculated as <span class="math">Var(X) = \sum_{i=1}^{n} (X_i - \bar(X))^2[/spoiler], where a bar above a variable means that it's the mean of that variable.
If you want to calculate the variance of something like <span class="math">X = Y + Z[/spoiler], <span class="math">Var(X) = Var(Y) + Var(Z) + 2Cov(Y,Z)[/spoiler]. You get this result by plugging in <span class="math">Y+Z[/spoiler] for <span class="math">X[/spoiler] in the original variance formula and working it out.
With three variables it becomes a bit messier. If I remember correctly, the variance would be in the form of <span class="math">Var(W) = 9Var(X) + 16Var(Y) + 36Var(Z) + 24Cov(X,Y) + 36Cov(X,Z) + 48Cov(Y,Z)[/spoiler], but you should double check that by plugging in your definition for W into the variance formula and simplifying it a bit.

Hope that helped.

The variance of a variable $X$ is calculated as $Var(X) = \sum_{i=1}^{n} (X_i - \bar(X))^2$, where a bar above a variable means that it's the mean of that variable.
If you want to calculate the variance of something like $X = Y + Z$, $Var(X) = Var(Y) + Var(Z) + 2Cov(Y,Z)$. You get this result by plugging in $Y+Z$ for $X$ in the original variance formula and working it out.
With three variables it becomes a bit messier. If I remember correctly, the variance would be in the form of $Var(W) = 9Var(X) + 16Var(Y) + 36Var(Z) + 24Cov(X,Y) + 36Cov(X,Z) + 48Cov(Y,Z)$, but you should double check that by plugging in your definition for W into the variance formula and simplifying it a bit.

Hope that helped.