My research is on the barriers that affect the willingness of persons with disabilities to engage in entrepreneurship.
My dependent (Willingness) and independent variables are all ordinal.
I have a sample size of less than 100.
What would be the most appropriate method of analyses (both qualitative and quantitative) to show the significant factors that determine Willingness.
Suggestions and comments would be really helpful.
Ordinal regression could be a potential method to see what are the storest predictors for Willingness. As a rule of thump, you would need 15 data points per independent variable in the model. However, your CIs get wider and your power to detect significant results get lower with more predictors/independent variables. A complex model with many independent variables might be out of the reach. In that case, you can consider the step-wise model selection.
Small sample sizes aren't useful as they do not allow statistically valid conclusions to be drawn. However if that is your non-negotiable constraint, then you might consider converting some of them into in-depth case studies for qualitative analysis. Oxford University's Bent Flyvbjerg is an expert on case study methodology and he is here on Researchgate. I discussed case study methodology in my dissertation methodology (attached). Hope that helps.
Thesis An analytical tool to aid the reflective selection of equity...
You could try t-tests or Analysis of Variance (ANOVA) or a Generalized Linear Model - you can read up on these and see if any work for your study.
Thank you very much for taking time to respond.
I will look into all your suggestions and see what would be best.
Your Independent variables can be nominal .Why ordinal ? Multiple regression can be used to compare the effectiveness of individual variables.
Wanny Eman, for people to offer you really helpful advice, I think you need to indicate how many people are in your sample. You say fewer than 100, but is that 99, or is it 3? :-)
I think you need to give a bit more information about the nature of your variables.
Small sample sizes typically between 5 and 30 sample. Is not advisable for quantitative approach. Use qualitative approach eg using focus group questioning may suitable.
The information provided by you is not complete. What do you mean by small sample? The other piece of information required is what you are required to do whether you want to conduct test of mean or test of relation. In case of mean difference there are non-parametric tests such as wilcoxon, Mann Whitney, Kruskal Wallis ANOVA. In case of relation PLS smart will work as it provides robust results even with the small sample.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size