How do I calculate my dissertation sample size?
For estimating a proportion in a large population, Cochran's formula is:
n = z² × p(1 − p) ÷ e²
- z = confidence level value (1.96 for 95% confidence)
- p = estimated population proportion (use 0.5 when unknown)
- e = margin of error (0.05 for ±5%)
Using these values gives approximately 384 responses for a large population. For a finite population, apply the finite population correction. For a population of 500 employees, the corrected sample size is approximately 217, consistent with the commonly used Krejcie and Morgan table. The appropriate sample size depends on the research objective, population and statistical analysis.
How many respondents do I need for regression analysis?
Green's (1991) commonly cited rules use m for the number of predictors. For the overall regression model, the rule is 50 + 8m. For individual predictors, it is 104 + m. With four predictors, these give 50 + 8(4) = 82 for the overall model and 104 + 4 = 108 for individual predictors. These are practical rules of thumb, not universal guarantees. A statistical power analysis is more precise and should be used when possible.
What is the difference between probability and non-probability sampling?
In probability sampling, participants have a known chance of selection. Examples include simple random, stratified and systematic sampling. It can support generalisation to the population when the sampling frame and design are appropriate. In non-probability sampling, selection does not use random selection; examples include convenience, purposive and snowball sampling. It may be appropriate when no complete sampling frame is available or when a specific group is needed, but findings have limits on how far they can be generalised. Explain and justify the approach you use.
Should I use validated scales or create my own questionnaire?
Established, validated scales have evidence supporting their measurement properties and can strengthen a study. Cite the original sources for the scales you use. You may adapt wording to fit your context, but justify modifications and consider validity and reliability again. Creating new questions may be appropriate when existing instruments do not measure your research construct adequately.
What makes a good survey question?
Ask about one issue at a time and use plain, neutral language. Avoid double-barrelled questions and leading, vague or ambiguous wording. Use consistent and balanced response scales. Put screening questions first and demographics near the end. Keep the questionnaire concise and relevant.
Why should I pilot my questionnaire?
A pilot can check question clarity, completion time, whether response options work, online form usability, and potential confusion or missing response options. Cronbach's alpha is a measure of internal consistency for a scale. A value of 0.70 is commonly used as an acceptable guideline in many contexts, following Nunnally (1978), but interpretation depends on context and is not an absolute rule. Use the pilot to identify and fix problems before the main data collection.