Hello!
I'm working on a dissertation of stress related sector among bus drivers. the objective is to explore factors that identify at which extend driving is stressful event among professional bus and coach drivers and explore the driving stress related behavior, willingness and openness to the change of those behavior among professional bus and coach drivers.
I used a likert scale questionnaire with 6 factors related to stress but each factor has at least 4 items.
which descriptive statistics should i compute, for example should I compute the mean at factor level or at item level? should i consider the distribution normally distributed or not? it is possible also that when you do cross tabulation all variable are not significantly related or there is a problem somewhere?
I would like advice step by steps of which analysis are suitable for this kind of situation.
Thank you
It seems you may be off base. The base regression is determined by the DV which appears to be likert unless I am mistaken because I can't follow your description. If it is correct you need ordinal logistic regression because the DV will be ordinal. It seems you are contemplating an ols regression which will be incorrect ir the DV is likert which is ordinal. I would delete the last full paragraph of your question which is largely incorrect for this case. I am attaching some notes that you will find helpful. If I have misunderstood you please redescribe the situation. With best wishes David Booth
BTW a sample of size 40 will not do for any contemplated regression which will not run with a sample of size 40 most likely. Best wishes, David Booth
Dear Nsengiyumva Olive Ladouce ,
I you designed the items yourself you should first test your factors for reliability. As your sample size is so small I would recommend running principal component analysis on each suspected factor. Please refer to my study guide on scale reliability with small samples.
After this it sounds like you have a single measure of stress (which David Eugene Booth calls your DV) and other factors which you want to test. The best technique is multiple linear regression (MLR) but you should start with scatter plots of your DV against each factor in turn. Field has a chapter on this technique which includes a discussion on sample size adequacy. It is a complex technique and I regret that I do not have a study guide on ResearchGate at present.
Essentially, your sample does seem to be very small in terms of Type II errors but if you have a strong "signal" in your data you might still find something. If the assumptions of MLR are not met then you will need to try a nonparametric approach. If you only had one factor this might be a Spearman correlation or a Chi Square test but I will leave it up to others to suggest alternatives with multiple factors.
Mean, median, mode, min, and max are insightful. Kruskal-Wallis is a good test to compare the factors.
- Badges
- Science topic
- Similar topics
- Biological Science
- Physiology
- Physiological Processes
- Stress