W6 Onwards Flashcards
(23 cards)
Chi-square is ONLY for which of the 4 kinds of data?
Nominal.
What does chi-square GOF (goodness-of-fit) do? On what condition?
Compares the observed counts (actual results) to the expected counts (expected results) IF the null hypothesis is true.
What does a chi-square TOI (test-of-independence) do? How is it different from GOF (goodness-of-fit)?
This is for two variables. Using an example, if we’re looking at whether there’s a significant difference between the number of males and females in Psych, then we use a GOF. But with two variables, i.e. how many males and females in psych versus in engineering, we use a TOI. They both compare observed versus expected counts.
Overall, chi-square helps us question whether what we observed is different from what we expected by _______ ? (In other words, if Ho is true.)
CHANCE alone. Is it due to chance or no? That’s why chi-square has p-values.
What does it mean if a p-value is significant in a chi-square test?
If a result is statistically significant, that means there is a relationship, i.e. the results are not due to chance alone. This is because we are assuming that under .05, there is only a 5% chance that we are making a Type I error.
What does it mean if a p-value is not significant on a chi-square test?
If it is not significant, i.e. above .05, then that means we have a chance of making a T1Error (rejecting Ho when it is true). It means there is no effect, i.e. the results ARE due to chance.
T-test is for how many sets of data MAX.? And what’s the condition for the sample size?
- Samples MUST be the same size.
What is a t-test?
The difference of 2 means divided by the standard error (SD divided by n squared).
What is between-subjects design? How is it done?
Looking at the mean between 2 different sets of people or 2 different groups. People are randomly assigned to either group. The bigger the sample, the likelier there is less inequality between groups.
What is a within-subjects design? How is it done? What is a possible issue?
1 group, measured twice.
T1 and T2 will be either control or intervention condition.
A possible problem is carry-over effect/ order effect, where people learn how to do the test better, or things might change over time.
What is a matched-samples design? What sample size is it useful for?
When people are grouped according to the condition that matches them best, i.e. when measuring whether coffee helps complete sudokus faster, people will be put in the group that drinks coffee a lot or the group that does lots of sudokus.
This is for smaller samples, with bigger ones there is less of a need to account for inequality.
What is a cross-over design?
It is both within and between-subjects design. Test both groups at T1, each one in a different condition. T2, reverse the conditions.
2 facts about the Standard Error:
A) it is the standard deviation of the sample means
B) larger samples have smaller standard errors
If Standard Error bars overlap on a bar graph, this suggests…
That the difference is not statistically significant.
What are 3 assumptions of the t-test?
A) That variance across groups is equal.
B) The data is normally distributed, i.e. m, median and mode all same
C) Data must be interval-level or higher
Related sample t-test is for _______ subject design. Independent sample t-test is for _______ subject design.
A) within-group
B) between-group
When is an ANOVA used? What does it do?
When you have more than 2 sets of data.
It compares sources of variance, between and within-group.
What letter statistic do we use in ANOVA?
F statistic or F-ratio
What is F?
The mean sum of squares between groups divided by the mean sum of squares within groups.
Rule: Large ______, small ______
Large F, small p: little variance within groups, large between. There is therefore little chance results are due to chance alone.
Post-hoc is ONLY done when:
The F-ratio is statistically significant.
What is a Bonferroni adjustment? What is the condition?
A post-hoc test that reveals where the difference is, as ANOVA cannot.
Bonferroni doesn’t work for more than 3 groups.
