IS 451 - Self made

what it is - the error you get if you just guess the average every time.

Think of it as - the “do nothing” baseline.

Analogy - a weather forecaster who says “70 degrees” every single day, because thats the average

SST - the lazy guesser

What it is - the error your model makes

Think of it as - your model’s score. Lower is better

Analogy - a forecaster who actually checks the clouds and the wind before guessing

SSE - your model

What it is : How much better your model is than just guessing the average.

the catch : it goes up ( or stays the same ) every time you add a variable, even a useless one.

Analogy : a teacher who gives you extra credit just for turning in more pages, even if the pages are blank

R squared , the “easy grader”

What it is : _____ , but with a penalty for each variable you add.

The rule: it only goes up if the new variable helps enough to cover its penalty

Analogy: a sports team that pays a fee for every player it adds. A great player is worth the fee. A bench - warmer just costs you.

Adjusted R - squared , the “fair grader”

_______ keep adding variables just because regular R squared goes up. it always will.

don’t

______ use adjusted r - squared to compare models with different number of variables.

do

A __________ tells you how much the prediction moves when one variable goes up by 1.

___________ = how much y (outcome) moves when one x (input) goes up by 1.

Coefficient

What it is : how easily luck alone could explain the result, if the true effect were zero.

Small ____ value ( below 0.05 ) : luck is a bad explanation. The effect is probably not zero.

Big _____ value ( above 0.05 ) : luck could explain it. We can’t tell if it’s zero or not.

Analogy. flipping a coin. Getting 6 heads out of 10 could easily be luck, so that’s a big p.
Getting 10 heads out of 10 is hard to blame luck, so that’s a small p. The coin is probably rigged.

P - Value

what it is : your x variables are highly correlated with each other. They move together.

the problem: the model can’t tell which one is doing the work.

Analogy: two singers on one mic. You hear the song fine, but you can’t tell who’s louder.

Multicollinearity

x correlated with y = good. that is a useful predictor.

x correlated with another x = ________________. that’s the problem.

multicollinearity.

_______________ can cause these issues


signs can flip. A variable that should be positive might show up negative.

p-values rise. variables that really matter can look “not significant”

Coefficients get shaky. small changes in the data can swing them a lot.

Multicollinearity