Stats, Ex. 3, Lec 1: 03/24/2014
T tests compare 2 different samples to see if they are significantly different or not.
We’re making a guess, the more people we have in a sample, the better we can make a guess. Makes it
unlikely that we’ll have 2 means that are that far apart due to chance.
Three different groups of children: At the end of treatment:
Usual care: Average of 1.86 diagnoses.
Standard manualized treatment: average of 1.54
Modular treatment: Average of 1.23
Modular seems to be working best. Fewest illness diagnoses.
Is this a significant difference?
Could do 3 independent measures Ttests.
Having too many Ttest conditions can greatly increase the chance of a type I
error. Type I error rate can be really high, we need to avoid this.
Ex. If alpha = 0.05
Test wise error rate= 0.05, but experimental wise error rate >0.05
We need a way to compare all these groups at the same time, how?
Use an analysis of variance, ANOVA.
Looking at variance: Want to see the different sources of change. From the example above, we get 3
If the curves are very narrow, there’s less variance, meaning each data set is doing a good job of
estimating each of their respective population means, meaning their difference is statistically significant.
If the curves are very wide and there’s a lot of overlap, there’s a lot of variance. There’s a
variance between the group means as well as variability within each group. The data could be spread out
as long as there’s isn’t too much overlap between the groups (the groups are different). How spread
out are the means relative to how the curves are spread out.
Comparing variance between conditions and the variance within conditions.
In ANOVA: in the example above, you have:
Independent variable: type of treatment. Dependent variable: # of diagnoses.
How many levels of independent variables do you have? 3