Unit 9

Question 1

Hypothesis

Answer 1

*Always in terms of parameters

Question 2

Null Hypothesis

Answer 2

H0 -> assumes existing claim, assumed to be true unless sample data is different ENOUGH than expected.

Reject the null when P<alpha.

H0 : P/Mew = #

___ IS ____

Question 3

Alternate Hypothesis

Answer 3

HA -> Null in rejected in favor of alternate hypothesis if sample data is different ENOUGH

HA : P/Mew </>/≠ #

___ is greater/less/not ____

If P ≠ , multiply by 2
If P >, -1
If P <, do nothing

Question 4

P-Value

Answer 4

The probability, assuming H0 is true, that your observed sample statistic would take a value as or more extreme

Smaller P = more evidence against H0

“If I take many many samples, (weird thing) will happen (P-value)% of the time”

Question 5

Alpha

Answer 5

Also called “level of significance”

Pre-determine decision point, stated with hypothesis. If P is below alpha, your result is “significant”

Reject the null when P < alpha

Question 6

Claim

Answer 6

Consists of 6 parts:

1. H0
2. H0 in context
3. HA
4. HA in context
5. Test type
6. Alpha

Question 7

Conditions (One-sample and two-sample situations)

Answer 7

INR

Question 8

Calculations

Answer 8

Stat - par / error = P (Reject the null if P<a)

Question 9

Context

Answer 9

Provide a nerdy answer and an answer in context

Question 10

Two-Sample Tests

Answer 10

Mostly the same, but the null hypothesis is always that there’s no difference between the groups, alternate hypothesis is that there is a difference or difference is less than/greater than parameter (read the question)

Stat - Par / sqrt. sigma1^2/n1 + sigma2^2/n2

REMOVE OUTLIERS

Question 11

Type 1 Error

Answer 11

(Alpha)

Stating something is going on, when actually nothing is going on

Question 12

Power

Answer 12

(Awesome!)

Something weird is going on, and you caught it

Question 13

Boring

Answer 13

Nothing is going on, and you suspected nothing

Question 14

Type 2 Error

Answer 14

(B, Beta)

Stating nothing is going on, when actually something is going on

Question 15

Error sentence structure

Answer 15

I conclude ___ but the truth is ___

Question 16

Choosing Alpha

Answer 16

If Type 1 Error is worse -> A = 0.01
If Type 2 Error is worse -> A = 0.1
Equally bad -> A =0.05

Why? You want type 1/2 errors to seesaw out, for example a type 2 error can’t be manipulated, but by raising the type 1 error the type 2 error will lower

Question 17

What if P is bigger than +/- on my Z-Chart?

Answer 17

Two methods:

1. Stat -> test -> test type -> put in information -> P should be given to you
2. If it’s less than 3, it’ll just continue getting smaller, so write P < .0002 (Remember to multiply or -1 accordingly)

Question 18

When do you use Chi (X^2) tests?

Answer 18

Used in categorical data, making it always proportional. used when there is 3 or more p-hats.

Question 19

Goodness of Fit test (GOF)

Answer 19

One row/column, single variable. Think: Is there a difference between observed (sampled) data and expected (population) data?

Stat –> list –> edit –> L1 = observed data, L2 = observed total * percentage of data
THEN
Stat –> tests –> X^2 GOF Test

Question 20

Test of Association

Answer 20

Multiple row/columns, multiple variables. Think: Is there an association between these two categorical variables?

2nd + x^-1 –> edit –> A
THEN
Stat –> tests –> x^2

You can go back to matrix to check your expected values (B)

Question 21

Degrees of Freedom (Chi-square)

Answer 21

(rows - 1) * (columns - 1)

If there is only one row/column, simply -1

Question 22

Conditions (Chi-square)

Answer 22

NORMALITY does NOT exist!!!

You can write:
Expected values are greater than 0
Independence met/not met
Randomness met/not met
Sample size sufficient

Question 23

Expected value equation

Answer 23

(Row * total) * (column * total)/(total * total)