Hi all,
I recently posted a question about how to conduct a convergent validity analysis for a self-developed self-esteem scale. I got some very useful responses! However, something I am stuck with is how to determine the sample size I need to conduct a convergent validity analysis between our self-esteem measure and the Rosenberg self-esteem scale. A lot of the quoted necessary sample sizes are larger than I expected, but they also assume that I would want to conduct a factor analysis. Yet, I do not plan to do so, as I do not expect our measure to exist of multiple factors, and the number of questions is quite small as well. Does anybody know what are rules of thumb for sample sizes in case your measurement loads on just one factor, and all you want to do is determine the level of correlation with another established measurement (the Rosenberg self-esteem scale)?
Thank you so much!!
Dear Sharon Schroen
I think you may use the rule of thumb that each item(question) requires from 10 to 20 respondents. However, there are other approaches for determining the minimum sample size which are more sophisticated as power analysis and monte carlo analysis, however, I am not sure if they are applicable in your context.
Best Regards,
Belal Edries
The most commonly used program for power analysis is G*power, but I am not sure whether it covers the analysis you want to do.
If the factorial analysis is by means of structural equations it is recommended to read the comments I referred to about the calculator for these models according to the number of latent and observable variables. We also refer to the hundreds of articles that reference its use.https://www.researchgate.net/post/Statistical_power_analysis_and_sample_size_required_to_conduct_Latent_Class_Growth_Analysis
Belal Edries thank you! I was thinking of using that rule of thumb as well, together with Gpower analysis. I am just a bit confused which correlation to use when I am to determine convergent validity. I can see multiple correlations included, but I think I should use the z-test --> two dependent Pearson's r (common index). To clarify, I aim to just correlate our self-esteem measure with the Rosenberg.
David L Morgan I have that programme, and I was indeed thinking of using it. I am just a bit confused which correlation to use when I am to determine convergent validity. I can see multiple correlations included, but I think I should use the z-test --> two dependent Pearson's r (common index). To clarify, I aim to just correlate our self-esteem measure with the Rosenberg.
Cristian Ramos-Vera Thank you for responding once more! I am actually not going to do a factor analysis, and that is why I am a bit confused, cause a lot of the sample size recommendations are based on that. I did indeed have read of the documents you have sent me and they were very useful!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size