week 9- inferential stats Flashcards
purposes of inferential statistics
- determine likelihood that study findings reflect actual population parameters versus chance
- test hypotheses about a population
hypothesis testing steps
- create null and alternative
- decide level of significance (alpha level)
- run the test (stats software)
- compare the probability (p value) against the alpha value
- make decision whether to reject the null or fail to reject the null
alternate hypothesis
what the researcher believes the outcome will be, directional or non-directional
null hypothesis
no difference exists between groups, researcher expects to reject the null
probability
- an event’s frequency in repeated trials, based on sampling error and theoretical distributions
- cannot prove alternate hypothesis but rather show support by rejecting the null
sampling error
- fluctuations between samples (approximation of the population)
- reduced with sample size
standard error of the mean
- statistics tend to fluctuate from one sample to another
- sampling distribution of the means is shaped like a normal distribution (central limit theorum)
- standard error is reduced by larger sample size
- smaller standard error = less variability b/w sample means
p-value
- provides evidence against the null hypothesis
- probability of getting a test statistic this extreme if our null hypothesis is true
- a specific area in the tail of probability (very unlikely to have occurred by chance)
- smaller p-value = stronger evidence
alpha or significance level
- usually 0.05 or 0.01
- indicates the there is a 1 or 5% chance that the results are a fluke
- purpose is to prevent a type 1 error
comparing probability value with alpha value
- if p-value < alpha value, results are statistically significant (reject the null)
- if p-value > alpha value, results are not statistically significant (fail to reject the null)
confidence intervals
- a way to express your conclusion as an interval with lower and upper bounds on same scale as original data collected
- smaller CI = more precise
- usually at a confidence coefficient of 95%
odds ratio
- used to determine the association between two variables
- determines the odds of having the outcome occur
- odds ratio of 5.5 indicates the likelihood of the outcome is 5.5 times greater than the comparison
odds ratio values
- OR > 1 indicates increased occurrence of an event
- OR < 1 indicates decreased occurrence of an event
- OR = 1 indicates the IV has no impact
type 1 error
- rejecting the null when it is actually true
- incorrectly accepting the alternate hypothesis
- consider reliability and validity of instrument
- level of significance = probability of making a type 1 error
type 2 error
- accepting a null hypothesis when it is false (failing to note a statistically significant difference b/w groups)
- may result from a small sample size
power analysis
- estimates the sample size needed to obtain a significant result
- type 2 errors can be reduced by doing a power analysis
- power of 0.8 is conventional standard (20% risk of making a type 2 error)
effect size
- magnitude of the effect/relationship
- larger effect size = greater effect/stronger relationship
- if the researcher expects a smaller effect size, you need a larger sample to demonstrate a difference
i.e koen’s D
koen’s D values
effect size value, ranges from 0-1.4
0.2 = small ES
0.5 = medium ES
0.8 < large ES
correlation coefficient values
0-0.2 = very weak/no relationship
0.2-0.4 = weak relationship
0.4-0.6 = moderate relationship
0.6-0.8 = strong relationship
0.8-1 = very strong relationship
positive vs negative correlation
a) positive = vary in the same direction
b) negative = vary in opposite directions
