Chapt. 13, research methods "Understanding research results: statistical inference"

### statistical significance

the difference between 2 (or more) means [results] is not likely to have occurred by chance. IV had effect on DV; difference not due to measurement or sampling error

### value for statistical significance

0.05. [any difference that is so large that it would occur by chance fewer than 5 times out of 100 (.05) is statistically significant]

### t-tests vs. F tests

t-tests are used to determine whether 2 scores differ by chance. F-tests are used to determine if 3 or more scores differ by chance

### what are F and t-tests ratios of?

a ratio of the variability between groups to the variability within each group [b/t grps:w/in each grp]

### How do you interpret the results of an F/t-test?

If the between groups variability is much larger than within groups variability, we would conclude our IV had an effect.

### Type I Error

**Say there is an effect when there actually isn't** Rejection of null hypothesis but null hypothesis is actually TRUE (we should not have rejected it).

### Type II Error

**say there isn't an effect when there actually is** Failure to reject the null hypothesis when we actually should have. type II= 1-Power

### Why are nonsignificant results hard to interpret?

There may have been no effect of our IV OR there could have been other reasons: manipulation may have failed, sample may have been biased, may have been a confounding variable, or effect of IV may have been so weak that we didn't have the POWER to detect it.

### null hypothesis

the population means are equal. samples came from the same population- the observed difference is due to random error.

### research hypothesis

population means are NOT equal. samples came from different populations; same population or no effect on the IV (no difference between the b/t and within groups)

### when is there possibility of a type I or a type II error?

Type I: when you say there is a relationship/effect; Type II: when you say there isn't a relationship/effect

### Things effecting power

1) Sample size (^=^ power) 2) strength of manipulation (and good operational defs) (maximize the between grps. variability); 3) reduce within groups variability

### Effect size

r-squared; what's the practical effect?; how meaningful a significant effect is; ex: if r= .6, then r2= .36 so 36% of "___" is accounted for by "_ _ _ _ _"