Hypothesis testing: L3-4 Flashcards
why use statistical tests?
how likely it is that we got our results by chance or if the results represent a real difference
- why use the t-distribution ?
- what does it take into account?
- we don’t know the population SD
- expected mean, measure of the standard error of the mean based on the sample
what do the DF change?
shape of the distribution
-> looks broader for lower df
-> normal distribution for large df
= cut-off varies depending on df
experimental design: one-sample design
- what is it
- advantages
- disadvantages
- one group values coming from different people
- used to compare group data to known values
- may not know population values
- may want to compare 2 groups to investigate change of behaviour over time
- may not know population values
experimental design: between-groups/ independent-measures design
- what is it
- advantages
- disadvantages
- 2 groups, values come from different people
- measurements are independent
- don’t have to worry about learning effects due to repeat exposure
- measurements are independent
- people in different groups might be v different -> need large sample sizes or counterbalance all factors that might influence results
- cannot study behaviour over time
- people in different groups might be v different -> need large sample sizes or counterbalance all factors that might influence results
experimental design: within-groups/ repeated-measure design
- what is it
- advantages
- disadvantages
- single group providing data for both conditions
- no differences b/groups
- study changes in behaviour over time
- test less people
- no differences b/groups
- measurements not independent -> need to calculate variance differently
- repeated exposure
- measurements not independent -> need to calculate variance differently
what do we assume about Ho?
it is true until we are sufficiently convinced that our result is “very unlikely” under this hypothesis
how to calculate est. standard error of the mean
standard deviation
_____________
sample size square root
- how to get standard deviation?
- how to get SS?
- take the square root of the variance
- all raw values - M squared and added together
what do we find to reject Ho?
t empirical > t critical
any value > 2.353 is what?
less likely than 5% to occur by chance
what would a non-directional hypothesis look like?
H1 = Our sample is drawn from a population μ>10 OR μ<10
some difference
For non-directional hypothesis what type of tests are used?
two-tailed test
-> .025 either end of distribution
what would a directional hypothesis look like?
H1 = Our sample is drawn from a population μ>10
higher/lower proportion
For directional hypothesis what type of tests are used?
one-tailed test
-> .05 one end of distribution
what can an independent-measures t-test do?
see whether 2 groups, exposed to different experimental conditions differ in a particular measure
independent-measures t-test formula
same as general structure, but now tests whether a difference between 2 means is significantly different from chance
What is the complication of this test & how is it calculated?
2 variances considered when calculating standard error of mean
-> pooled variance
how is pooled variance calculated?
summing up the summed squared differences (SS) from each condition & dividing by the sum of the df
= only works if both groups = same n
whats the paired-samples t-test formula?
same structure of the t-test but now tests whether the average difference score is significantly different from chance
what cant the t-test tell us?
- how meaningful the effects really are
- > bc its highly dependant on the sample size
effect size measures:
- cohen’s d gives?
- small, medium & large effect sizes
- estimate of effect size independent of the sample size
- -0.2
- 0.5
- 0.8
effect size measures:
- variance explained r2 gives
- small, medium & large effect sizes
- estimate effect size NOT independent of the sample size
- 0.01
- 0.09
- 0.25
- 0.01
What’s a confidence interval?
-> why set it?
- range of values centered around a sample statistic
- should be near to the corresponding population parameter
- > to make sure our population mean is contained somewhere within that interval (e.g. 95% confidence interval -> make sure our pop.mean was contained within in 95%
Assumptions to consider before deciding whether to run a t-test (3)
- observations must be independent (no systematic biases)
- populations must be normal
- if comparing 2 populations, samples must have equal variances