Both within-subject and between-subject designs have advantages and disadvantages. Are results different for each design type? How can I choose the appropriate experimental design type?
Hello Asma,
Just an example. It may help you to "transfer" this to your current research.
1. You are going to provide treatment for depression and you want to know whether it is statistically significant. Prior intervention you measure "level" of depression (time 1). Then you provide intervention and after it is finished you measure "level" of depression again. You are hoping to reduce depression severity significantly. (Paired sample (also called within subject) t test would be viable option if you have two mean scores for comparison).
2. Based on the same sample you just want to know if "level" of depression is higher for women or men. You "cut" your sample across gender and obtain two independent groups (women and men). If you have two means you can use independent sample t test also called sometimes between subject t test).
Hope it helps a bit
If you are interested in this, search literature for ' Mixed design'. It is possible to use a mixed design, which contains a between-subject factor and a within-subject factor. Such a design would consist of both a within-subject variable test and a between-subject variable test. Thomas Tullis and William Albert for instance mention a mixed design which uses genger as a beween-subject factor in a study with three trials distrubuted over time (within-subject factor) in the book Measuring the User Experience, 2nd edition.
This could also help:
http://www.stat.cmu.edu/~hseltman/309/Book/chapter15.pdf
http://www.bristol.ac.uk/cmm/learning/multilevel-models/
http://www.ltcc.ac.uk/courses/REML_and_LMM_Notes.php
http://www.bodowinter.com/tutorial/bw_LME_tutorial2.pdf
http://www.ats.ucla.edu/stat/mult_pkg/glmm.htm
https://www.statistics.ma.tum.de/fileadmin/w00bdb/www/czado/lec10.pdf
http://www.uic.edu/classes/bstt/bstt513/GLMM.pdf
Hello Asma,
Just an example. It may help you to "transfer" this to your current research.
1. You are going to provide treatment for depression and you want to know whether it is statistically significant. Prior intervention you measure "level" of depression (time 1). Then you provide intervention and after it is finished you measure "level" of depression again. You are hoping to reduce depression severity significantly. (Paired sample (also called within subject) t test would be viable option if you have two mean scores for comparison).
2. Based on the same sample you just want to know if "level" of depression is higher for women or men. You "cut" your sample across gender and obtain two independent groups (women and men). If you have two means you can use independent sample t test also called sometimes between subject t test).
Hope it helps a bit
Can you give a little better idea about the experiment you are dealing with? A type of design I have used, called a Gage R&R design, tests between and within subject variability. The design I used is nested.
Asma: Oui, c'est tout à fait possible. Techniquement, aucun problème. Or, il y a des situations o\u l'un ou l'autre des modèles (inter vs intra) peut s'appliquer (les postulats de base sont respectés) et vous pouvez alors faire les deux types d'analyses et rapporter les résultats pour les deux. Je vais tenter de retrouver un exemple tiré de mes propres analyses (il y a 20 ans!).
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology