## MSc Research Methods Week 4

**Note: this post is aimed as a memory mnemonic for me not you (I’m studying an MSc in Digital Marketing )**

**Week 4 – Sample Design and the MRS Code of Conduct**

The sample should represent the larger group of interest.

The major considerations in designing a sample = who, where, when and how (…do you interview)

The key goal in sampling is that you select a sample that is as representative as possible.

**The 5 Step Sampling Process:**

**Step 1. Define the Population**- Specify the elements, sampling units, extent & time.

**Step 2. Identify the Sampling Frame**- Where will your respondents be found? Basically identify the ‘sampling frame’

**Step 3. Decide on the Sample Size**- How many respondents will you use? And there are different approaches (arbitrary, conventional, cost basis, statistical analysis, confidence interval)

**Step 4. Select a Sampling Procedure**- Probability or Non-probability?
- Probability = every member of population has known probability of selection. Survey results can be projected to the total population. Only used in quantitative research. Essential for causal research.
- Non-probability = Sample selection based on judgement, used in both qual. and quant. Research and particularly appropriate for qualitative research & industrial research. Probably the one I will be doing as I will not have a full list of the population.

**Non-Probability Sampling **

**Note: You do not always have to do one or the other method; you can combine them… **

**Judgement Sampling is one reasonably good option**

This is based on selecting elements via “expert” opinion of worth to project, researcher draws, representative sample, based on judgment and particularly for test markets & product testing & B2B. **Problem**: selection bias

Another good option is **Quota Sampling (non random)**

– ID of subgroups that need to be represented

– quotas established for population sub-groups

– by judgement not probability

– interviewers given sampling areas

– Interviewers interview set no. of respondents

– per population category

– to build sample identical to population

Problems: Possible selection bias, non response bias, bias within quotas

**Snowball Sampling**

– selection of additional respondents based on referrals

- for low incidence or rare populations

** **Problems: bias & reluctance to give referrals

**Step 5. Physically Select the Sample**- How will you actually select the respondents?

**Sources of Error**

**Sampling Error**

- where sample value does not reflect underlying population
- easier to correct where measurable
- decreases as sample size increases

**Non-Sampling Error**

- errors occurring in MR process except sampling error e.g. faulty problem definition, defective population definition, poor questionnaire design, interviewer bias, data analysis factors, etc)
- more difficult to detect & measure
- do not decrease with sample size

**Ethical Issues**

Got to make sure I adhere to the Market Research Society (MRS = mrs.org.uk)) code of conduct i.e. need to be transparent, confidential (e.g. often do not ask for names to keep the information confidential and that names will not be used in the research), objective, honest.