Stats I Final Pitt 2

used to compare 2 means where subjects have been measured twice

• - subject is measured twice on a DV pre and post treatment (aka pre-post design)
• subject is measured on a DV at different levels of an IV simultaneously (DV = SAT, IV = math, verbal)
• -subject can be more than one person (DV = happiness, IV = spouses (husband, wife))

• Normality
• No Outliers
• Independence of subjects

• the difference score of a DV (error term) should be normally distributed
• tested using Shapiro-Wilk

• a research question can still be answered by using an alternative test that does NOT assume normality
• a non-parametric alternative to the paired-sample t-test is Wilcoxon signed rank test

• has an undue influence on a statistic
• assumption is checked by examining histogram and Q-Q plot

• design considerdation
• one subject score is not influenced by another subject in study
• cannot be tested using statistics

df= N-1

doesn't depend on the number of scores but rather on the number of subjects

• interval estimation
• accounts for uncertainty of error

Order effect

Carryover effects

Test effects

• when the outcome of the study is affected by the order in which the treatments were prepared.
• Two most common types: carryover and test effects

effects of one treatment do not end prior to the administration of another treatment, so the effects observed in the later treatment include residual effects of the former

ex: testing new pill but not waiting long enough between dosages

fatigue and practice effects

effects due to repeated testing of participants over time

ex: taking the same test twice in a short period of time

common solution to an order effect

the order of treatments is often counterbalanced so that some participants get treatment in order but other get a second order so that order effects can be assessed.

Reduced error: variance within subjects is typically smaller than between subjects (smaller sampling error, more power)

Number of subjects: paired-samples t-test requires smaller number of subjects to achieve the same power than the independent-samples t-test