Inferential statistics

Question 1

whats null hypothesis

Answer 1

saying that resluts are due to chance

Question 2

whats alternative/experimental hypothesis

Answer 2

saying that results are due to experimental effect, not chance.

Question 3

whats the logic behind null hypothesis testing (3 steps)

Answer 3

• prob that HO is right (hence results are due to chance factors ie. random sampling variability) follows a normal curve.
• alpha: defines the very low probability of the HO being correct (hence results being due to exp variables)
• region of rejection/critical region: if p value of NHST falls here, we reject HO and say its statistically significant.

Question 4

differentiate between one tailed and two tailed hypothesis tests

Answer 4

two tailed hypothesis tests are non directional - hence two tails at either end, alpha is split in two
one tailed hypothesis test is directional - hence one tail, equal to alpha

Question 5

describe one sample Z test

Answer 5

involves calculation of standard error: σM=σ/(√n)=(population SD)/√(sample size)

Then calculating Z score: z=(M-μ)/σM

Question 6

describe one sample T test

Answer 6

involves calcuating standard error, but with sample SD replacing mean SD

SM=S/(√n)=(sample SD)/√(sample size)

then we calculate T score: t=(M-μ)/SM

requires normal distribution and interval/ ratio data

Question 7

describe independent samples T test

Answer 7

when theres 2 samples, and they’re both independent of each other

requires normal distributino, interval/ratio data, homogeinty of samples

Question 8

describe paired samples T test

Answer 8

two samples but are paired in some way.

requires normal distribution and interval/ratio data

Question 9

why does sample size conofund effect size

Answer 9

because its used in all NHST’s, which generates p value; and p value is therefore not only due to effect size but due to sample size

Question 10

describe cohen’s d, and ranges of scores

Answer 10

Cohen’s d = |(mean distance)/SD|
>0.5 = small effect size
0.5-0.8 = medium effect size
>0.8 = large effect size

Question 11

describe type one error

Answer 11

false positive: incorrectly rejecting the null hypothesis

= alpha

Question 12

describe type two error:

Answer 12

false negative: incorrectly accepting the nullh hypothesis

= beta

Question 13

describe power

Answer 13

1-beta

when we correctly reject the null hypothesis

Question 14

what is good power?

Answer 14

0.8

Question 15

ways of increasing power?

Answer 15

increasing alpha (but this increases type I error)
increasing effect size (eg. exposure to treatment)
increasing sample size
using within participants design