Comparison of continuous variables between two groups Flashcards
What does it mean for a sample to be independent?
A sample is independent if the elements (e.g. people) are drawn/sampled such that the inclusion of one observation and outcome does not affect the inclusion of another or its outcome
What does paired data mean?
Matched/paired samples
May be repeated observation of the same person Measurements over time: paired data
What does no effect mean?
Often considered as no difference between the groups
Hence, a statistically significant result implies a difference
What are the first things to do when I have a data?
Simple summaries and checks:
Before beginning any rigorous analysis always check the data Spot issues in the data
Identify potential violations of assumptions
Check you have the data you think you have
Check if you have more data, which may affect your question
Pictures:
Plots are important for investigating data and presenting findings
Can also be useful for spotting data issues
What re the assumptions leading to a t-test?
Comparing means of two populations using two samples - one from each population
Assumptions:
-Known (z-test) or unknown (t-test) population variances?
-For unpaired samples, and if unknown population variances, are they equal (equal/unequal)?
-Normally distributed continuous outcomes
-Observations within each population are independent
-Are observations between populations independent (unpaired) or dependent (paired)?
What are two estimators of test statistic? And when should they be used?
-Satterthwaite’s approximation
-Welchs approximation
When to use either approximation? Further assumptions and technical details. Both are only valid for sufficiently large sample sizes (in total or per group?)
Which summary function naturally shows skewness and kurtosis?
describeBy
Have to load “psych”, using library function
What is a box plot good for?
Box plots are used to show distributions of numeric data values, especially when you want to compare them between multiple groups.
What do the Satterthwaite’s (Welch) variance look like?
Variance equal is false. Otherwise, R takes it as normal.
Paired data?
Must have equal number of observations in each sample Observations may be matched in a number of ways, depending on the question
How to specify paired t-test in R?
Paired = True
Page 28 session 5 slide
What axis should the after data be in a paired t-test?
Y-axis