What oh what, where can you be,
An appropriate model that describes me;
And now thanks to An Intro to Computational Data Analysis for Biology,
I know where to start, so let’s go, please bear with me.
The first thing to do is to ask what is my predictor?
What response does it predict, and what is its shape toward the future?
I skip over money and also alcohol,
as the latter causes over-dispersion, and of the former I have none-at-all.
If I’d like to go and compare my model to the rest,
just so I can prove once and for all I’m the best;
I must choose something common to everyone with which I can rhyme,
oh wait, that’s perfect, I’ll just go and use Time.
As my response I will choose “Happy Times/Day”,
It’s error Poisson Distributed… which is always fun to say;
So i go back and sample across time my life,
but there are always some gradients that may cause some strife.
My replicates are days, but I’m not a morning person,
so I randomly sample across 24 hours, just to be certain;
Happiness builds and diminishes across seasons,
so Temporal autocorrelation will be needed to account for those reasons.
So now here I have my life’s model, YAY!
But it’s really no fun without more to say;
So a-priori to gathering all of my data,
I outlined other variables to look at which may influence MX and Beta.
Incidence of approval from advisors of sorts,
Leisure time spent in skis, cleats, or swim shorts;
While these two should lead to happiness increased,
Long problem sets… ehem stats homework… will probably provide the least.
So now I go and fit my data to some math,
As happiness isn’t exponential or bounded, I assume linearity down my path.
family= poisson (link=”log”), data=My_Life) should be fine.
But before I evaluate, I look at correlations,
just to be sure there aren’t funky palpitations.
Uh oh, there one is, I’m gonna have a fit,
It’s Approval and Homework… I’ll never hear the end of it.
So I add in another variable to the equation,
“Approval:Homework” will now help explain the relation.
So I run my diagnostics, looking at residuals vs fitted,
Q-Q, Scale-Location, Cooks Distance, Residuals Leveraged.
When everything checks out, I look at my fit,
I evaluate an ANOVA, Coefficients after it.
I’m using and alpha of 0.05 even though sometimes it’s rubbish,
I need to conform to NIH guidelines so that I may publish.
My happiness is positively influenced by Leisure and Approval,
Also Approval’s interaction with Homework, but negatively upon its removal.
Apparently Time has no affect, I’ve been happy all along,
exemplified by constant proclivity to write rhyme/song.
So now I can say I’ve finally reached my goal,
of finding my model and evaluating it as a whole.
And though I’ve done so through end-rhyme poetry,
It’s a weekly reflection, so don’t hold it against me.
It’s been fun taking the course, and though I’m drowning in work,
the knowledge I’ve gained has been quite a perk.
Even with all I’ve learned, statistics still seems quite daunting,
not nearly the same mathematical meadow in which you, Jarrett, seem jaunting.
I do have more confidence, and I’m not talking about intervals,
when discussing my analysis and reasoning and residuals.
So here’s my big thanks to Biol 697,
It came just in time as I’m nearing the end of data collection.