Sample size and sampling procedure
This study applied both random and non-random sampling strategies. In selecting the study site, simple random sampling technique was applied whereby the names of the three Isiolo sub-counties were written on pieces of paper then folded several times and put in a container, which was shaken and the researcher picked one out. Three villages were selected purposively with the help of the local provincial administration who assisted the researcher.
To get a representative sample size, this study used a formula used by Mugenda and Mugenda (2003).
The sample size was determined as follows: n = z
2
pq/d
2
where
n = the desired sample size (if the target population is greater than 10,000)
z = the standard normal deviate at the required confidence level
p = the proportion in the target population estimated to have characteristics being measured
q = 1 − p
d = the level of statistical significance set
Since the target population is less than 10,000, the final sample estimate (nf) was calculated as follows:
$$ \mathrm{n}\mathrm{f} = n $$
$$ 1\kern0.5em +\kern0.5em n/N $$
where
nf = the desired sample size (when the population is less than 10, 000)
n = the desired sample size (when the population is more than 10,000)
N = the estimate of the population size (Mugenda and Mugenda 2003)
When the population is more than 10,000 households, 384 of them are recommended as the desired sample size (Mugenda and Mugenda 2003: 43). The accessible population in this study was 4,000 households.
Mugenda and Mugenda recommend the formula
$$ \mathrm{n}\mathrm{f}\kern0.5em =\kern0.5em \frac{n}{1+\frac{n}{N}} $$
to be used to calculate the sample size.
According to the above formula,
nf = the desired sample size when the population is less than 10,000
n = the desired sample when the population is more than 10,000
N = the estimate of the population size
Using the above formula, the sample size is
$$ \mathrm{n}\mathrm{f}\kern0.5em =\kern0.5em \frac{384}{1+\frac{384}{4,000}}\kern0.5em =\kern0.5em 350 $$
To cater for those households that would decline to participate or dropped out during the process of investigation, the study proposed a sample size of 400.
A total of 400 households were interviewed. These consisted of 206 men and 194 women. The proportion of men was more than that of women because men were the ones who were readily available.
Systematic sampling was applied to select households for the interview.
The sampling interval was determined by the equation given below.
$$ \mathrm{Sampling}\kern0.5em \mathrm{interval}\kern0.5em =\kern0.5em \frac{n}{N} $$
where
n = the required sample size
N = the population size
n = 400
N = 4,000
$$ \mathrm{Sampling}\kern0.5em \mathrm{interval}\kern0.5em =\kern0.5em \frac{400}{4,000}\kern0.5em =\kern0.5em \frac{1}{10}\kern0.5em \left(i.e.\kern0.5em 1\kern0.5em in\kern0.5em 10\right) $$
Microsoft Excel FUNCTION = RANDBETWEEN (1, 10) was used to select a random starting number for the first household to be included in the sample, which happened to be number 8.
The eighth household was from village 1 since the households were assigned numbers starting from village 1, 2 and then 3.
Data collection methods
This study was cross-sectional in nature, and both qualitative and quantitative methods were employed. Secondary data was collected through a literature review, while primary data was gathered in household interviews, in focus group discussions and from key informants.
Qualitative data were analysed according to the themes in the research objectives. Quantitative data were analysed using the SPSS software.