(2004; 98 pages)
5.6 Sample size
We now have to determine our sample size. It is a widespread belief among researchers that the bigger the sample, the better the study becomes. This is not necessarily true. In general it is much better to increase the accuracy of data collection (for example, by improving the training of interviewers or by better pretesting of the data-collection tools) than to increase the sample size after a certain point.
In qualitative studies the aim is not to be representative of the population. The validity, meaningfulness and insights generated from such studies have more to do with the information richness of the cases selected, and the analytical qualities of the researcher than with the sample size. There are no rules for sample size in qualitative research. It depends on what one wants to know, the purpose of the study and practical factors. Often qualitative researchers refer to the redundancy criterion: that is when no new information is forthcoming from new sampled units, stop collecting data. One can also use pragmatic criteria in defining sample size, considering the amount of time it costs to do and transcribe the interviews and the number of sub-groups from which one will select respondents. A qualitative study with 40 informants is a relatively large study. Generally qualitative comparative studies have at least 10 informants per group.
In quantitative studies, as a general rule we can say that the desirable sample size is determined by the expected variation in the data: the more varied the data are, the larger the sample size we will need to attain the same level of accuracy. You need to consult a statistician, who can usually make precise calculations to determine the desirable sample size. Examples of such calculations follow below. For descriptive studies, we cannot say more than that the sample size needs to be large enough to reflect important variations in the population, but small enough to allow for intensive study methods. You should aim for at least 30 people in each group of interest. The EPI-Info 6.04 software includes an easy to use sample-size calculator.
In a study on reasons for non-use of oral rehydration therapy, you may decide to interview two categories of informants (non-users and users), and start with 20 to 30 interviews per category. This number could be increased if the data obtained for each category do not indicate a certain trend or if results are conflicting. The eventual sample size is usually a compromise between what is desirable and what is feasible.
A quantitative study should aim to quantify well-defined variables, for example, the proportion of under-five-year-olds treated with oral rehydration therapy. Sample size calculations are based on estimates of what these proportions are likely to be (informed guess or results of previous surveys). These estimates are made before selecting a sample. For a simple random sample the table can be used to determine the required sample size.
Example: The aim of the survey is to measure the proportion of people going to the village shop. Although there is no clear information on this, it is assumed that 40% of the people would go to a village shop. This is taken as the preliminary estimate, i.e. a population of 0.4 goes to the village shop. From table 3, it is seen that the desirable sample size for a proportion of 0.4 is 145.
Sampling size calculations for multi-stage sampling are more complicated. It is best to consult a statistician. Statistical advice is also needed to define sample sizes for comparative studies (such as those done in evaluation studies when experimental groups are compared with control groups), where one wants to test differences between two groups. The desirable sample size can usually be calculated, with some assistance, if the researcher is able to make a rough estimate of the outcome of the study, and is clear about its main objectives and variables.
Table 3. Sample size for a simple random sample^{1}
ESTIMATED PROPORTION |
DESIRABLE SAMPLE SIZE* |
ESTIMATED PROPORTION |
0.05 |
420 |
0.95 |
0.10 |
325 |
0.90 |
0.15 |
290 |
0.85 |
0.20 |
255 |
0.80 |
0.25 |
225 |
0.75 |
0.30 |
195 |
0.70 |
0.35 |
170 |
0.65 |
0.40 |
145 |
0.60 |
0.45 |
120 |
0.55 |
0.50 |
100 |
0.50 |
The desirable sample size is given in the middle (second) column. The table is entered using either the left (first) column or the right (third) column depending on whether the estimated proportion is less than or greater than 0.5
* For the information of survey specialists: In this table the estimated S.E./p gradually increases from 0.10 for p = 0.5 to 0.21 for p = 0.05
^{1} Source: Lutz. W. 1982, ibid.
The feasible sample size is determined by the availability of resources:
- time
- human resources
- transport
- money.
Remember that if people are to be interviewed in their homes, it is often more time-consuming to go and trace the people than to actually do the interview. In addition, remember that resources are not only needed to collect the information, but also to analyse it! If many variables are included in the study (which is usually the case in an exploratory type of study) the sample size should be relatively small to avoid problems during analysis. If one has few variables, one can afford to have a larger sample.
The following general rules may help to determine the desirable sample size of any given study:
• the desired sample size depends on the rates one expects for key variables.
• the desirable sample size also depends on the expected variation in the data (of the most important variables): the more varied the data, the larger the sample size one would need to attain the same level of accuracy. For descriptive studies it is important that the sample size is large enough to reflect important variations in the population, but small enough to allow for intensive study methods.
• the desirable sample size also depends on the number of cells one will have in the cross-tabulations required to analyse the results. A rough guideline is to have at least 20 to 30 study units per cell.
References
Hardon A et al (2001). Analysis of qualitative data. In: Applied health research manual: Anthropology of health care. Amsterdam, Het Spinhuis.
Hudelson PM (1994). Qualitative research for health programmes. Geneva, World Health Organization. WHO/MNH/PSF/94.3.Rev.l.
Lutz W (1982). Sampling: how to select people, households, places to study community health, 3rd ed. International Epidemiological Association. Edinburgh.
Varkevisser CM, Pathmanathan I, Brownlee A. (1992). Designing and implementing health systems research projects. Volume 2: Geneva, Health Sciences Division of the International Development Research Centre and World Health Organization.