What up kid?
You say stat is rough, hurtin your head.
When you got stuck, you should a asked me instead.
ahhh Just ask Mac.
Think back how you begun.
Is hypothesis test your first step, son.
Don't reject the null, fail to reject the null either.
Until you find whether the alpha value is higher.
Quit flippin.
Talkin about stat is so tough.
Make sure you listen to the lesson while we're telling you stuff.
Do you wanna compare one group to another
Or perhaps one group to some known number?
If your sample size is at 30 or more n >= 30
you best be testing this with the z score.
If you remember these words it ain't no terror.
Observed minus mean over standard error. obs - mean
------------
std dev
When you check your z too for the critical sum...
Degrees of freedom equals n minus one. x - _
i x
What test. What test. z = ----------------------
i sigma
What test. What test.
R, F, T or B
Chi squared from A to Z
Now if you're workin with a sample small in size
then the z test will not suffice.
You're gonna have to use an alternative theory
try t test to answer your query
but if you wanna get a correct solution
first you must assume a normal distribution.
equal variance is a needed feature...
but if not use a sample width procedure
you'll have to
and we told you to remember this forever
observed minus mean over standard error
if your samples are better matched
a new course of action must be hatched
calculate
our apologies if we're monotonous
but when your DV is dichotomous
if your p value's less than point oh five p < .05
if you're looking to be innovative
and you got variables that are qualitative
conduct chi square and you'll do well
just make sure you expect five in every cell
sometimes two variables in the ranges
one can predict how the other changes
if you got this distinct impression
you're in need of some linear regression
To minimize error in your prognostication
calculate the least squares equation
This can be accomplished with ?????
or you can just uses some matrices
Beta nought is where you cross the y B B
beta one is the slope of the line 0 1
R T and F all indicate whether your model can hold its weight
if the SSE shows you're not too wrong,
there's a real good chance that your model's strong.
if you really wanna know how well it should fare
hope you can find a large adjusted R squared
is ...
not ...
but what if you got more than one predictor
use multiple regression and you'll be the victor
Get your evidence for the model validity
just be aware of multicollinearity
Which model is best,
which is robust
which model leaves the others in the dust
So when this class is ancient history
and you're into research or industry
Remember these tips and you won't get stuck
This is real s**t we're not makin this up...