Thursday, May 19, 2011

Well to make myself feel better I think I will list all the concepts I've learned in the last two years and the stuff I need to learn hopefully over the summer.  This is just a partial list.  I will add to it as I think of things.

Still don't understand/have yet to try:
MCMC
Markov Chains
Priors and posteriors
Conjugate priors
Bayesian statistics
Network construction
Measure theory
Lebesgue integration
Bootstrapping
Jackknife
Ridge regression
Lasso

Learned/tried
LDU + SVD + Spectral + Cholesky Decompositions
Riemann sums/integrals
closed open sets
proof by induction/contradiction/contraposition
introductory topology
field axioms
limsup
liminf
metric spaces
compact 
complete
what a ring is
convergence of series
convergence of sets
uniform convergence
delta method
taylor series expansion and the remainder term
finding eigenvalues/eigenvectors
affine transformations
gradient/jacobian 
hessian
positive definite/negative definite
hermitian
lagrange multipliers
delta-epsilon proof method
lipschitz continuity

simple linear regression
ordinal logistic regression
ANOVA
ANCOVA
MANOVA
hazard functions 
survival function
Kaplan-Meier
dealing with left and right censored data
types of distributions
method of moments
characteristic functions
sufficient statistic
exponential family 
multilinearity
bootstrap
biased 
consistent estimators
UMVUE
BLUE
BLUP
pseudoinverse
oracle estimator
shrinkage


correlation structures
autoregressive correlation
longitudinal analysis
LME
GLM
GLMM
GAM
GEE
weighted least squares
iterative weighted least squares
newton-raphson
fisher scoring
principle components analysis
dose dependence curves
maximum likelihood
REML
quasi-likelihood
likelihood ratio
AIC 
BIC

fisher information
score equations
overdispersion
dose response 
link function
cross validation
classification
mixture models
asymptotic convergence
log parallel assays
smoothing 
cubic splines
kernel smoothing
local polynomial regression
wald test

non-parametric inference
semi-parametric inference
wilcoxon-rank sum test
factor analysis
mahalanobis distance
person, deviance, standardized, studentized residuals
deviance table
stepwise forward/backward/both model selection
hotelling's t2









I didn't pass. Well I didn't fail either, but I still feel pretty bad. I worked as hard as I could have worked, but it didn't make much of a difference.

It was the first time I latexed a whole paper, I really used the stepwise regression in R which apparently I still didn't understand well enough, the first time I did PCA in R I think, and the first time I used leverage values and cook's distances in a report or understood them.

I guess some of those things are pretty basic to undergraduate level statistics. I know for a fact though when I took regression 108 in the summer we did not learn about leverage points or cook's distances, AND we used minitab. We didn't even do a stepwise regression except on paper...

And then over the summer when I worked with Sandy I remember she helped me with the model selection and ran the stepwise regression for me in SAS. The problem with the exam was that a lot of the questions that Burman asked me were not ones I ever learned in class, or they were ones I should have learned in his class I should say. Except he didn't teach it. Maybe it's because I took it with him over the summer and it covered less content or maybe it's just something you pick up later on on your own, but it's definitely not something explicitly taught to me.

I tend to retain things better when I'm actually taught them. I feel sad about being the only one who didn't pass my oral, but I can't really blame myself that much. The description for the exam said to write a publishable data analysis and do an oral defense. As far as I'm concerned, there were no problems with the quality of my data analysis, all the problems were conceptual. If I had known that was the kind of questions I was going to get, I would have studied and reviewed my 108 materials instead of spending so much time doing the data analysis and trying new methods.....

I still feel bad though. No matter how much I learn, people always seem to expect more from me. I just don't think I can assimilate material that fast. I can make a list of every new thing I've learned in the last two years that would be pages long... I just think there's only so much I can cram into my brain at a time. And it's not all that fair because almost everyone has had 4 extra years to cram into their brain this stuff, so naturally they know more. What I'm really getting is a masters and bachelors degree in statistics in 2 years, without having taken a single math class in college. At times when people are frustrated with me, I just want them to realize that, I wouldn't expect them to be a math major and get a masters degree in cell and developmental biology in two years.

Wednesday, May 18, 2011

The paper I co-authored last summer is already published.

http://www.ncbi.nlm.nih.gov/pubmed/21493108


Looks like this summer will be another busy one. I meant to leave time to prepare for my classes/exams next year, but I will be GSRing at 100% again.

My MS oral exam is tomorrow. After about 30+ hours latexing. I am really exhausted. I think I can finally sleep after it's over. I just haven't really slept at all it seems, either cause my brain wont shut down, or I've been working late at night... I'm really nervous too. Hopefully it will go well.

Tuesday, May 10, 2011

Some thoughts

Sometimes being yourself is hard if yourself, is not within 1 standard deviation of normal... Actually there is no such thing as normal.  If there is even a bell curve, it's distribution is most likely multi-modal, not to mention multivariate.  The idea is though, that it's easier for people who are close to those medians to be themselves.  Just like it's easier for someone with a normal looking face to go into public than someone with a deformed looking face.  It just makes sense.  There's less to criticize.

So being yourself for different people can take different levels of courage I think.  Which is sometimes why it seems easier to try to blend in.  But there's always that secret fear that if they knew who/what you were like on the inside, then they wouldn't want to hang out with you anymore.  So it's easier, but in some ways, it's not, because you're never at ease.

What does this have to do with anything?  It doesn't really.  They are mostly some brain rambles.  Leftovers of thoughts from events that have impacted me recently.  Recently, a question was brought up about whether or not killing Osama Bin Laden was a Christian thing to do.  The group decided that it was not.  But it made me think how in various different groups people might come to entirely different conclusions.  There's no such thing as a Christian stance on something.  Sure, there is a conservative stance, a conservative Christian stance, a liberal stance, a progressive Christian stance, and many other stances.  And most likely people who have these stances hang out mostly with other people who share their views.  Again, people tend to feel more comfortable around other people like themselves.  But if the price of fitting in is adherence to a idea or creed, I don't think anyone is truly at ease.  

What I mean to say is, I think the best kind of organizations are the ones in which people share different opinions.  And on any one topic there is no one correct answer.  Life is, and always will be complicated.  It shouldn't be about being right, but merely thinking.  Having a view and defending it should be just as valuable as getting an A.  Which is why I feel so often like I'm failing at school as I'm rushing to turn in something that I know is the right answer, but the only reason I know it's the right answer is because I went to office hours and had it explained to me, not because that's the solution I came up with using my own logic and thought process.  

That's my two cents for today.