CHAPTER 4: RESEARCH AND COMMERCIALIZATION POLICY IN ZIMBABWE
5.6 SAMPLE AND SAMPLING
For the purposes of the study, all samples were drawn from the respective populations using Krejcie and Morgan’s (1970) model, acknowledged and confirmed by The Research Advisors (2006), constructed from the formula:
“s = X 2NP (1− P) ÷ d 2 (N −1) + X 2P (1− P)”
Where:
s = required sample size,
X2 = the table value of chi-square for 1 degree of freedom at the desired confidence level (3.841),
N = the population size.
P = the population proportion (assumed to be .50 since this would provide the maximum sample size),
d = the degree of accuracy expressed as a proportion (.05) (Krejcie & Morgan, 1970:607).
Table 5.2 below shows an extract of the table for determining sample size from a given population (Krejcie and Morgan, 1970) courtesy of The Research Advisors (2006). Using this pre-determined samples table, the researcher acquired samples based on the respective populations available for the study. All samples were drawn at 95% confidence levels, at a 5.0% Margin of Error, which are the generally acceptable levels in research (The Research Advisors, 2006).
Table 5.2: An Extract Table for Determining Sample Size from a Given Population Required Sample Size
Source: The Research Advisors (2006:2)
Of the twenty research institutes, 19 had been scheduled to be purposively sampled in view of Krejcie and Morgan’s (1970) table for establishing sample size from a known or prescribed population, as shown above. However, a sample of 15 institutes was used due to the fact that the remaining institutes were not at liberty to allow their institutions to be used for research purposes, thus permission was not granted to include such in the study. These came mainly from the agricultural sector and they included some of SIRDC’s research institutes (Biotechnology Research Institute, Environmental Sciences Institute, Building Technology Institute, Food and Biomedical Technology Institute, Geo-Info Remote Sensing Institute, Production Engineering Institute, Electronics and Communications Institute, Metallurgical Research Institute), Harare Institute of Technology, Cotton Research Institute, Coffee Research Station, Agronomy Research Institute, Pedstock Investments, Energy Technology Institute and Taisek Engineering. These represented about 65% of the research institutes in Zimbabwe, which was deemed a fair representation of the total sample thus rendering the results of this study generalizable and reliable.
Out the 800 “data-based” customers for the research institutes, a sample of 260 respondents was drawn using the same model (Krejcie & Morgan, 1970) and this represented 32.5% of the population.
Regarding research institutes, interviews were held with six purposively sampled R & D and
commercialization and / or business development-related management staff, while a sample total of 94 respondents were drawn based of varying institutes’ databases as follows:
Table 5.3: Staff Members’ Sample Sizes per Research Institute
Institute Population
(N)
Sample (n)
SIRDC institutes (x 9 institutes) 65 56
Harare Institute of Technology 8 8
Cotton Research Institute 5 5
Coffee Research Station 10 10
Agronomy Research Institute 5 5
Pedstock Investments 6 6
Taisek Engineering 4 4
Totals 103 94
Source: Developed by the Researcher
As can be noted from Table 5.3 above, for populations of less than 50 cases or subjects, they were selected in their entirety as also supported by Henry (1990), who is apparently more of an opponent of probability sampling for such cases. Henry (1990) proposes that one must “collect data on the entire population as the influence of a single extreme case on subsequent statistical analyses is more pronounced than for larger samples” (quoted in Saunders et al., 2007: 208).
5.6.1.1 Location and Geographical Dispersion of the Research Institutes
Figure 5.1: Geographical Dispersion of the Research Institutes Used For the Study Source: Google Maps, 2013
Although the majority of the research institutes were concentrated in Region 2 (Harare, the capital city), the nation was fairly represented in the sense that the study also engaged institutes from regions 3, 4, 5, 7 and 10. These institutes represented 65% of the research institutes in Zimbabwe, which was deemed a fair representation of the total sample, thus rendering the results of this study generalizable and reliable. The researcher engaged only those institutes where written consent was given, permitting use of their institutes for the purposes of this study. Most of the institutes preferred anonymity, especially regarding results presentation and analysis, and this was well observed by the researcher.
5.6.2 Sampling: Techniques and Procedures
Amongst many other methodology scholars, Trobia (2008) and Miller and Salkind (2002) point out two basic sampling methods, that is, “probability or representative sampling and non-probability or judgmental sampling”. Due to the mixed nature of the study’s methodology, the researcher used both probability and non-probability techniques. While the use of probability techniques allowed the researcher to statistically estimate the population characteristics from the sample, non-probability sampling techniques indeed made it possible to make some statistical inferences in to the characteristics of the population as guided by Saunders et al. (2009) and Trobia (2008).
The study utilized simple random sampling to select the sample of 260 customers out of a population of 800 respondents. Simple random sampling was also used in selecting a sample of 56 of 65 SIRDC’s employees. With the increasing popularity and usage of various programs, the researcher used an Online Random Number Generator (Haahr, 2010), courtesy of random.org website. Simple random sampling was deemed necessary because the study subjects possessed almost similar characteristics and the geographical dispersion was somewhat concentrated (Saunders et al., 2007). A number was assigned to each customer and / or employee from the relevant databases from 1 to 800 (for customers) and from 1 to 56 (for SIRDC institute’s employees). Parameters were set in terms of the “minimum” and “maximum” numbers, followed by clicking the “generate” function repeatedly until the samples of predetermined size were drawn. Thus every subject of the population had an equal and independent chance of being incorporated or picked by the Random Number Generator.
Using this online program ensured very limited researcher intervention thus reducing bias, at the same time ensuring that the desired samples could be selected with little effort (Miller & Salkind, 2002).
Figure 5.1 below shows a snapshot view of the Random Number Generator taken from the random.org website.
True Random Number Generator
Min: 1 Max: 800
Generate Result: 100
Powered by RANDOM.ORG
Figure 5.2: Online Random Number Generator Snapshot (Source: random.org, 2014; Haahr, 2010).