Week 10 (t-tests) Flashcards
When do we use a t test?
When we want to compare up to two sample means
What are the type of t tests? (3)
One-sample
Independent-samples
Paired-samples
What are the general assumptions of t-tests?
Data are interval or ratio DV
Data is normally distributed
Scores for an independent-samples test are independent
Equal variances (for independent samples test)
What is a one-sample t-test?
Determines whether the sample mean is statistically different from a known or hypothesised population mean
What occurs in a one-sample t-test?
The test variable is compared against a test value which is a known or hypothesised population mean
How do we calculate the t-value for a one-sample t-test?
t = (mean from sample - expected mean)/estimated standard error
When do we use an independent-samples t-test
Used when we can’t to compare means from two groups
How do we calculate a t-value from and independent samples t test?
t = (X1-X2)/standard error of difference between means
X=
Sample means
Does sample size affect t distribution?
Yes it varies with sample size
How do we summarise an independent t-test?
t(n+n-2)= X, p= X, 95% CI [X,Y], d = x
How do we find the confidence intervals for an independent t test?
Jamovi doesn’t calculate so use excel effect worksheet
Explain confidence intervals overlap in independent t-tests?
If CI’s do not overlap more that 25% of their total length, the difference is statistically significant
What is an expectation for paired-samples t-test?
Scores are independent
Pairs of scores will be correlated (however not a requirement)
What is the null hypothesis for a paired-samples t test?
The mean of difference scores = 0
What calculations occur in a paired-samples t-test?
There are two scores for each participant, one from each condition
Calculate difference scores for all participants
Calculate the mean difference score
How do we calculate cohens d for a paired samples t-test?
Doesn’t take into consideration the correlation of two conditions so use effect size worksheet to calculate a better d
What does the Shapiro-wilk test analyse?
Checks whether the data are normally distributed
If test is significant, then there is an issue with the normality of the data
If not significant, assume meets normality assumption (p values)
How to calculate degrees of freedom for a paired samples t test?
N-1
Should you report standard error in a paired samples t test?
Not normally reported so convert to SD by multiplying it by the square room of df+1
Which t test do we use for a between subjects design?
An independent samples test
Which t test do we use for a within subjects design?
A paired samples t test
How do we test the assumption of homogeneity of variance in independent samples t tests?
Levene’s test
When is the homogeneity of variance satisfied?
When the variances of the DV at the two levels of IV are equal or nearly equal
What happens when levene’s test is not significant?
If the p value is not statistically significant it indicates the standard deviations for the two groups do not differ significantly
The data meets the assumption of normality
If levene’s test is not significant which statistics do we report?
Students t stats
What happens when levene’s test is significant?
If the p value is statistically significant it indicates the SD’s differ for the two groups and the assumption of homogeneity of variance has been violated
Which statistics should be reported when the levene’s test is significant?
Stats from Welch’s t
What is the difference between student’s t and Welch’s t?
Identical in terms of t statistic
Welch’s test compensates for the violation of homogeneity of variance by applying an adjustment to the degrees of freedom, making it smaller.
Why does making the degrees of freedom smaller change the differences between the means?
The smaller the df for a t test, the larger the value of t test required for the difference between the means to be significant
Which t-test require categorical IV’s?
Paired and independent t tests
Welch’s test and errors?
More conservative, decreasing the risk of type 1 error but increases the risk of type 2 error.
Confidence intervals will be wider for this test