Sampling and Data Collection. In October and November of 2016 and after ethical approval was granted for the study through Emory University’s Institutional Review Board, the SurveyMonkey® link was distributed nationally through the ACNM email listserv and through social media to CNMs and CMs certified since May, 2011. The sampling frame was the ACNM email listserv cohort that included 1,474 members. These members were contacted twice with reminders after the initial email. The link was also posted on social media several times, and contacts were encouraged to share the link. A total sample of 269 CNMs (no CMs) responded to the survey. Due to the snowball sampling distribution of the link through social media (in addition to distribution through the ACNM email listserv), the response rate to the survey distribution is unknown. Data from those individuals who graduated before May, 2011 or who were not employed at the time of the study were removed, and data from surveys that had mostly complete responses for background, education, and workforce sections were retained, resulting in a possible 244 CNM responses for each item. The link’s introductory page included survey information, disclosures, and contact information for the principal investigator. Clicking to progress past the introductory page indicated informed consent. Survey completion rate was 93%, although not all respondents who completed the survey responded to every question. Survey results were downloaded from SurveyMonkey® to IBM SPSS statistical software version 24. Quantitative data were initially reviewed for normality, outliers, and implausible values. Missing data were examined for type, extent, and presence of bias. Categorical variables were checked for sparse cells and regrouped as needed. Skewed results from 5-point Likert scale questions were dichotomized such that “strongly agree” and “agree” responses are grouped separately from “strongly disagree,” “disagree,” and “neutral” responses. Statistical analysis, with alpha set at 0.05, was performed for descriptive data, correlations, t-tests or Chi squared tests, and regression parameter estimates for the research questions associated with study aims. To minimize multicollinearity and select the most parsimonious final models, tolerance and variance inflation factor were assessed, and variable selection methods were used within each regression to identify significant predictors. Respondents were from 42 U.S. states, the District of Columbia, 1 U.S. territory, and 1 in...
Sampling and Data Collection. In October and November of 2017 and after ethical approval was granted for the study through Emory University’s Institutional Review Board, the SurveyMonkey® link was distributed nationally through the ACNM email listserv and through Facebook to CNMs and CMs certified within 5 years. The ACNM email listserv cohort included 1,474 members, who were contacted with an initial email plus two reminders. The link was also posted on social media several times, and contacts were encouraged to share the link. A total sample of 269 CNMs (no CMs) responded to the survey. Due to the snowball sampling distribution of the link through social media, the response rate to the survey distribution is unknown. Data cleaning involved removal of those respondents who graduated before May 2011 or who were not employed at the time of the study. Data from surveys that had mostly complete responses for background, education, and workforce sections were retained, resulting in a possible 244 CNM responses for each item.
Sampling and Data Collection. It is important to facilitate robust comparisons between the interim and final evaluation that a consistent sampling and data collection approach is used. Therefore, in line with the interim evaluation, we recommend using Capibus, which is the Ipsos MORI weekly omnibus survey. Capibus interviews a nationally representative sample of 2,000 British adults aged 15+ each week. The survey would only be asked of those who are sole or joint decision makers (expected to be around 1600 adults). A summary of the Capibus approach is included below and a more detailed account is in Appendix B. Ipsos MORI face-to-face omnibus Face-to-face in-home survey among 2,000 adults aged 15+ Runs weekly providing responsiveness and flexibility to fieldwork dates Nationally and regionally representative of adults across Great Britain Up to 180 sampling points randomly selected each week CACI ACORN used to set quota controls specific to each interviewer location Findings for GB overall are statistically accurate to +/- 1 percentage point Very latest multi- media CAPI software allows us to play show reels to respondents and display shots on- screen Strict quality control with 1 in 10 interviews back-checked by telephone Findings from a survey have a confidence interval, or margin of error, associated with them due to the fact a sample of the population is being interviewed, and not everyone. The approximate confidence intervals for various sample sizes related to this survey are shown in the table below. The confidence interval is widest at a finding of 50% and narrows the nearer we get to absolutes of 0 or 100%. This table shows the confidence interval at the 95% confidence level, which means we can be 95% certain that the result lies somewhere within the margin of error indicated by the confidence interval. Strictly speaking the tolerances shown here apply only to random samples; in practice good quality quota sampling has been found to be as accurate. We have estimated the number of respondents likely to fall into the customer group based on the interim evaluation. It balances our expectation that CERT (across its lifetime) will have reached more households than previously, but that we will also be able to validate professionally installed customers against EST data to take account of over claim (see section 3.4.1.3 on analysis). The estimate of Priority Group Customers is also based on the interim evaluation, while the Super Priority Group is based on a calculati...
Sampling and Data Collection. The primary goal of a research is to get representative data. To achieve this, we need to either enumerate the whole population or select a representative sample. Such that the researcher can study the smaller group and produce accurate generalization about the larger population (Xxxxxx, 2003). Determination of the sample is not an easy task. It is subjected to several factors. Such factors include the type of sample, statistic to be applied, homogeneity of the population, time, money, and personnel availability for the study (Xxxxxxxxx and Xxxxxxxxx, 2002). Care must be taken to ensure the sample is not biased, i.e. some types of study objects (like people) are not more likely to be sampled than others. In our case we have not done so. Brokers and rural traders are underrepresented, or downstream actors are overrepresented. It is possible to compensate for that through appropriate weighting of the data. In Tanzania there exists no documentation of farmers and traders or middlemen/brokers dealing with pigeonpea. In order to get the chain of pigeonpea marketing, we needed to start from the source with the farmers who produce the crop and follow the value chain downstream from them. Farmers who sell at the farm gate invariably rely on itinerant assemblers/brokers who visit their villages during the harvesting season. Information about these assemblers/brokers could therefore be collected at the village level. In our case, we collected this information from farmers‟ self-help groups in a random sample of villages drawn from a list of all pigeonpea-producing villages in Babati district (see appendix 1) Information about the self-help groups that are found in the villages was provided by DALDO – Babati, non-governmental organizations and Gendi farmers cooperative office in Babati. This enabled us to assemble the list of farmers groups and their original village (see the list in appendix 1). By using the snowballing sampling procedure which is common for social network studies, a random seed sample from the farmers groups dealing with pigeonpea was selected in targeted villages in Babati District. The groups were asked to identify the four most important pigeonpea brokers and traders who were operating in their villages. (This constraint was rarely binding – in most cases the groups listed all the assemblers/brokers they knew who were operating in their villages) (See appendix 1, table 8.1). In the terminology of this procedure, the groups were "primary responde...
