hypothesis testing and statistical significance Flashcards
confidence intervals
how confident can we be that the results in our sample represent the population
it is an interval estimate of where the population mean might lie in the population.- e.g. we are 95% certain that the population mean lies between 8 and 10
in psychology we use 95% confidence interval based on SD OF 2(1.96).
standard error
standard deviation divided by square root of the sample size.
you need the standard error to calculate your confidence interval.
confidence interval calculation
you times the standard error by 1.96 (sd)
then you plus or minus it from the mean- and this is your range that u are 95 certain the population mean lies between.
issues with standard error
smaller samples have larger confidence intervals
so larger the sample the better the estimate of the population
hypothesis- alternative
H1
There is a difference/relationship between the variables
population mean from the 2 groups are not equal
directional- important you have previous research to back this up
can be causal- specific causal influence- only used if you are doing a controlled experiment
or non causal-suggests particular characteristics of behaviour without reference to causation.
can be directional or non directional- if enough previous evidence we can do directional.
directional= one tailed test
non directional= two tailed tests
hypothesis- null
H0
when we do statistical testing we translate it to the null hypothesis
there is no difference/relationship
population groups are equal
hypothesis and falsifiability
we need to be able to test hypotheses generated by a theory that prove the theory incorrect
p value
5%= 0.05
you ask spss of the probability of finding such a difference if there was no relationship in the population (if the null was true)
if P value is less than 0.05 we can reject the null hypothesis
Inferential stats
tell us about the population- draws conclusions about the population based on what we observed in our sample e.g. is the results we saw in the sample likely going to happen in the population- we make inferences by checking the p value
lots of statistical tests are used
need specific assumptions before you do a certain test
T tests (parametric)
William Gosset (1908)- referred to as 'student t test' difference between two groups/means- can be on 2 separate groups/ one group on 2 occasions/ whether one group compares to a specific mean (we don't cover this one).
parametric vs non parametric
PARAMETRIC- based on population parameter (e.g. mean or SD)- they assume your data plots in a normal distribution- so you need to screen your data beforehand- more power= ability to find a effect/difference/relationship when using smaller sample sizes
Main overview of assumptions= .ratio or interval of DV
.population should be normally distributed
.variation of populations should be equal(only valid when doing tests comparing 2 means)
.no outliers/extreme scores
NON-PARAMETRIC- distribution free- don’t make any assumptions about the distribution- these are less sensitive meaning you need larger sample sizes (less what we call power)
T TEST assumption checking
.must be interval or ratio
.’reasonably’ normally distributed
.if comparing 2 different groups the variances across populations must be equal. (Levens test checks this)
.no outliers/extreme scores
degrees of freedom in a T test
one sample t test= N-1
related t test= N-1
unrelated t test= (N-1)+(N-1)
one tailed or 2 tailed ?
2 tailed preferred- non directional/null
spss will automatically run 2 tailed
to do 1 tailed in spss u divide p value by 2
Types of T tests
- within participants/repeated measures/paired sample t test- advantage of this is it has higher power as natural differences are controlled for
2.