Hi all,
I had a fairly general question regarding analyse and sample sizes. Essentially I am wondering what kind of psychometric analyses are possible with smaller samples (around 100) on a 28 item questionnaire using ordinal data (strongly agree - strongly disagree).
I was hoping to do CFAs/EFAs, however my worry is that 100 respondents won't be enough. Aside from coefficient alphas, skewnes, kurtosis, are there any other psychometrics that could be carried out? The purpose of the tool is to al confirm that the items fall into well established scales.
CFA & EFC with a small sample has low power. You can use temporary stability with test-retest technique. Also convergence and discriminance validity.
Hi Darko Giacomini ;
In some situtaion, sample size of 50 could be sufficient. Take look this articles:
So may be you can use EFA for construct validity.
Article Exploratory Factor Analysis With Small Sample Sizes
-
Article Recommended Sample Size for Conducting Exploratory Factor An...
Besides multiplying 10 bt the maximum number of indicators of a latent variable, use the GPower Analysis to determine sample size.
Alejandro Ros-Gálvez , Abdullah Faruk Kılıç , Rodrigo Moreta-Herrera ,
Thank you all for the insights!
Bootstrap may be suitable for this type of situations.
Best wishes.
İbrahim.
Rasch style analyses can work with a sample size of around 100. This isn't recommended for high stakes testing, but it may give you a picture. For details on sample sizes and Rasch analyses see: https://www.rasch.org/rmt/rmt74m.htm
You'll want to examine if your underlying instrument structure is suitable first though.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size