Unit 10: Inference- NHST, CI, Effect size

Question 1

nhst: null hypothesis significance testing

Answer 1

research question

Question 2

H1

Answer 2

alternate hypothesis

Question 3

H0

Answer 3

null hypothesis (we only test the null: reject or fail to reject)

Question 4

what does it mean by significant

Question 5

null hypothesis is true but you reject it- type 1 or 2 error

Answer 5

type 1

Question 6

null hypothesis is false but you dont reject it- type 1 or 2 error

Answer 6

type 2

Question 7

alpha a

Answer 7

probability of rejecting the null when it is true (type 1 error)

typically set at .05

Question 8

beta b

Answer 8

Probability of not rejecting the null when the alternative is true (i.e.,
Type II error)

Question 9

Power

Answer 9

• Ability to detect significant results when they are truly there
• Power = 1 - β

Question 10

tradeoff between alpha and beta

Answer 10

Question 11

p value

Answer 11

The probability of you getting that observed statistic by chance if the null is true

if it is less than alpha, we reject the null hypothesis,
and the result is significant

• Typically, alpha is set at .05
• So the result is “significant” if p < .05

Question 12

The level of significance p

Answer 12

The level of significance p (or a) specifies how rare results must be in order to provide evidence of rejection of H0
* This means the result is likely not due to chance (measurement error)
* The low threshold for statistical significance is p=0.05
* Other level of significance used in practice are 0.1, 0.01, and 0.001

Question 13

One-Tail versus Two-Tail

Answer 13

A one-tailed test looks for an “increase” or “decrease” in the parameter whereas a two-tailed test looks for a “change” (could be increase or decrease) in the parameter.

We go with two-tails

Question 14

If im asking Is my sample less than the population?
* Is my sample more than the population?

Answer 14

• uni-directional
• One-tail
• In practice, always two-tails because it’s “more
  conservative and stringent”
• In a sense, it’s like we really don’t want to falsely convict an innocent person

Question 15

z test

Answer 15

Basic purpose: tells us about the underlying distribution the sample is drawn from. (is the sample mean likely to represent the population with which it is being compared

• Because we know the distribution of sample means will be normal,
  no matter what, we can see where our sample mean falls on the z-
  distribution (standard normal distribution)
• Do this by using the population mean (μ) and SD (σ )
• Null: H0 distribution
• If we find that our sample lies in
  an extreme tail of the H0 distribution,
  then we argue it may come from a
  different distribution

Question 16

standard error

Answer 16

standard deviation of population/sqrt of sample size

Question 17

z test vs z score

Answer 17

Question 18

example of using z test

Answer 18

Question 19

confidence interval

Answer 19

Confidence interval is a range of values within which it
can be stated with reasonable confidence the
population parameter lies

• Calculated from sample data
• Tells us the precision of a point estimate
• We can set the level of confidence. Typically, it’s 95%,
  or sometimes 99%.
• You see it reported as 95% CI or 99% CI

Question 20

Statistical significance

Answer 20

means that an outcome isn’t likely to
have occurred by chance

Not statistically significant = representative = a random occurrence = occurs
within 95% of outcomes in H0 distribution
Statistically significant = not representative = a non-random outcome = occurs in the extreme 5% of outcomes- likely from alternative distribution
* An outcome or difference that exceeds a z score value of +/- 1.96 on the normal distribution is statistically significant.

• If value of the z-statistic is <= -critical value (e.g., -1.96)
• Or value of the z-statistic is >= +critical value (e.g., 1.96)
• Then, you reject H0.