So you've decided to do a randomized design with a post test only
Here is a simple model of the design:
-------------------------
Rt -> I -> Po
-------------------------
Rc -> -> Po
-------------------------
Rt=Randomized treatment group
Rc=Randomized comparison group
I=Intervention
Po=Post-test
Analysis of Designs with Post Test Only
The easiest way to understand and analyze designs with post test only is :
I. T-Test:
A T-test is a ratio that calculates , [Difference between groups]/[Variability between groups]. It is a signal to noise ratio..
T= [MEAN(treatment)-MEAN(comparison)]/Standard Error(Trt.Comp)
Here is an example of a T-Test shown in MINITAB : __________________________________________________________________________________________________________
Two sample T for Comparison group vs treatment group
Comparison group: N=100
Mean= 0.143
Std Dev=0.986
SEmean=0.099
Treatment group:
N=100
Mean= 0.030
Stdev=0.990
Standard Error (mean)=0.099
________________________________________________________________________________________
*95% CI for mu Comp - mu treat: ( -0.163, 0.388)
T-Test mu Comp = mu treat (vs not =): T= 0.80 P=0.42 DF= 197
________________________________________________________________________________________________________
The results of this T-test are not significant. We know this because the confidence interval (indicated by * above) spans zero (0) CI: (-.163,.388). This is the easiest way to tell whether your test is significant or not.
For your information, designs with post test only can also be analyzed using a :
2) Linear Regression
MODEL:
Outcome=Constant1 + (constant2)Zi + error
Z=0 for comparison group, Z=1 for control group
OR
3) Analysis of Variance (ANOVA)
MODEL
outcome= U + Ai+ Error
U=Population mean on all observations
Ai=effect of treatment (either treatment or no treatment)
More information on ANOVA and Linear regression can be found in Applied Linear Statistical Models, by Neter, Kutner, Nachshem, and Wasserman.
The easiest and/or most versatile packages for T-tests, ANOVA and Linear Regression are :
- 1) MINITAB-very user friendly-point & click
2)SAS-sorry, you have to write your own code for this one
3)SPSS-again,point & click
Please Email me with any information
or comments about this web page
4/9/97