Difference Flashcards
Paired-samples t-test
2 experimental conditions and same participants took part in both. Tests diff bwtn 2 means
Independent t-test
2 experimental conditions and diff participants assigned to each one. Tests diff bwtn 2 means
Sample from same pop then means will be roughly
Equal
If standard error is small we assume
Samples to have similar means = accurate reflection of pop
Signal-to-noise ratio
Variance explained by the model divided by variance that the model can’t explain. Effect/error
Variance sum law
Variance of a difference btwn 2 independent variables is = to sum of their variances
Standard error
Tells how much variability there is in this statistic across sample distribution from same pop. Take SD divide by square root of N
Sampling distribution
The distribution of possible values if a given statistic that we could expect to get from a given pop
Dividing standard error does 2 things:
Standardises average difference between conditions
Contrasting the diff bwtn means against the diff that we expect to get
If difference btwn samples is large and standard error of diff small
Confident our difference is not by chance
t-statistic
Ratio of the systematic variation in the experiment to the unsystematic variation.
Parametric test
Test that requires data from one of the large catalogue of distributions that statisticians have described.
t-test assumptions
Based on normal distribution.
Differences btwn 2 or more means
t-tests, ANOVA, MANOVA
Relationship models
Correlation, regression
Descriptive stats
Mode median mean. SD. Describing data
Inferential stats
Draw inferences about populations from samples
Z score
Standardise scores relative to sample. Standardising scores. Calculate raw scores. Correspond to units of standard error
If conclude that sample mean has come from a pop with a given mean…
It has a relatively low z score (closer to middle)
High prob of occurring by chance in the pop
Retain null hypothesis
1.96
Is the point of rejection in a 2 tailed test
Directional hypothesis - must argue…
If good research evidence then you can use directional hypothesis.
Alpha levels
.05 test result would occur 5% or less of time by chance
Non directional (2tailed) test alpha levels
1.96 (.025 in each tail)
Directional (1tailed) test
1.67 (.05 in one tail only)
If z >1.96 p must be <.05 meaning
Z is sign - reject null
Z-test
Based on zscore formula but using the formula for the sampling distribution of means. Sample mean / pop mean - standard error
Z (1.67) is < critical z (1.96) WHICH MEANS…
Retain null
Test statistic greater than critical value (ie z) then…
Significant effect
T-distribution when large (large sample) is…
Normal distribution
T-distribution is small (small sample) below 30
Flatter than normal distribution. Harder to detect a difference!!
Have to be ……… Than critical t for there to be a difference
BIGGER
One sample t-test
Data from a single sample - u want to know whether the mean of the pop from which the sample comes from is the same as some hypothesised mean
Critical t depends on…
The sample size
Sampling variation
Extent to which a statistic (z, t) varies in samples taken from same pop
What does kolmogorov-Smirnoff test measure?
Wether a distribution of scores is significantly did from normal distribution
Significant result from kolmogorov-Smirnoff test inidicates
Deviation from normality. Affected by size of sample though!
Which alpha level is more conservative?
.01
Most accepted level of probability…
.05
What does kolmogorov-Smirnoff test measure?
Wether a distribution of scores is significantly did from normal distribution
Significant result from kolmogorov-Smirnoff test inidicates
Deviation from normality. Affected by size of sample though!
Which alpha level is more conservative?
.01
Most accepted level of probability…
.05
