Annotation and coding of transcript
BUS : Processes and trade items from the fishing business
EVE: Change in behaviour and other events
MET: Method of fishing
PEP: Actors/ participants (People, animals, community
Coding help in developing relationship between concepts, time sequence of the fishing business and the competing fishing methods.
Coding indicates trade opportunities such as high quality meat, bones and carcase and hides spelt out in question 1.
Change in fishing methods over time is possible through coding from nets being left out in the bay and being floated to beach, to gun in Q2 and a regulated fishery in Q3. Competition has occurred affecting the respondent’s business prospects.
The changes brought by competition makes the speaker nostalgic of the pre 1967 era. The speaker thinks the dugong behavior was influenced to change by reduction in feed, intelligence of the fish and overfishing.
The respondent seems to respect the dugongs at the expense of other humans as seen by the use of “them” referring to fish and “indigenous” referring to other people. Not any form of association to other people or groups in the responses shows an unsympathetic person who just want to make a living from the dugongs.
Case study design of WMPRA
1. Sampling error
To reduce sampling error, sampling frames were created from maps showing all households in each village by the local residents and checked several times. However the use of local residents in itself though they know the area to come up with the correct maps, they may not be technically equipped. They can be biased in the creation of maps and boundaries.
The 341 households used in the survey is too small a number to generalise the research findings given the size of the target population spanning over 230km2 . The further reduction of the households from the initial 341 to 123 for logistical reasons further compromised the research finding. However to reduce sampling error a systematic random sampling was used.
The follow up interviews on the 30 fishers also is too small a number to rely on for generalisation of the findings. It is expected that the use of the various tools complement each other and improve the nature of the data collected and generalisations.
2. Sampling related errors
In this study this type of error emanated from the decision to sample the whole Watamu village. The decision resulted in an inaccurate sampling frame. Also associated with such is the non responsiveness.
The poverty wealth ranking (PWR) where households were ranked from the most to the least wealthy compounded the sampling related errors as PWR was not standard. For example the poorest in creek villages included skipping a meal or two per day, lows levels of education and lack of material assets and housing problems. By contrast in Watamu even the poorest had 3 meals per day and there were no reports of individuals lacking a house.
To improve on the poverty wealth ranking accuracy, the rankings were carried out three times per village to cross check people’s perception of wealth and placement of individual household.
3. Data collection error
Flaws in the administration of the research instruments such as personal interviews, observations and questionnaires resulted in data collection errors as follows:
Problems from interview arise for misunderstanding from the interviewee or respondents, memory problems from the respondents where they cannot remember in detail what has happened historically in a long time. The uneducated creek villagers did not keep any records on sales to confirm their income. Poorly worded questions in an interviews can cause problems in the interpretation and responses. The way the question is asked by the interviewer affects the outcome of the research if they don’t have the necessary skills and practice. Some participants did not understand Italian or English and efforts to circumvent this by engaging interpreters could possibly result in misinterpretation.
In trying to reduce data collection error through questionnaires, a household questionnaire was administered at four different points of the year reflecting distinct tourism seasons. However the required information was too much, and demanded too much time from the respondents. Failure to pay attention to detail when preparing the questionnaire result in grammatical and formatting errors which can influence the respondents’ first impression and attitude in answering the questionnaire(Bryman, 2016). The questionnaire structure can also affect the responses if questions are not ordered correctly. Also the low level of literacy result in some subjects not participating. All the above have the effect on the research findings. To improve on response the respondents remained anonymous and this enabled them to freely express themselves.
Personal observations which confirms that there was higher wealth in Watamu than creek villagers need trained observers to avoid observation errors and bias. Errors also emanated in the huge number of parameters to be observed.
Survey gave snapshot view of who had been engaged in fishing and or tourism over the year. However livelihood trajectories (LTJ) revealed that livelihood choices fluctuate over a much larger time scale than a year.
Lastly with regards to all the above, poor recording of information result in the outcome of the research being biased and unreliable.
4. Data processing Error
These are errors which arise from the management of collected data. Common among them is coding error where information collected is wrongly or poorly coded (Bryman, 2016). The transcribing of LTJ and entry of data into Atlas.ti for coding could result in errors. (706 words)
a) In depth household interviews were undertaken to collect data. This method helped in bonding the researcher and the respondents. The creation of trust further improved the rapport and the participants felt comfortable and were very cooperative. In depth interviews allowed for probing and prompting which enables the research to collect as much information from the respondents as possible. The interviewees feel very important with the personal touch since the sample was small and thus the researcher collected maximum information on the feelings, attitudes and perceptions regarding community forest management.
Interviewees are not influenced by others is a group and high response rates are achieved. The use of respondents’ own words avoid any misinterpretation of the responses. Interviewers can also monitor change in tone and other non verbal communications. The respondent complement of 80 is workable with since in depth interviews need fewer participants to glean useful information.
b) Stratified random sample of 40 house hold was used in each Barangay was taken for a total of 80 household interviews. The list of household was obtained from the provincial management office and verified to avoid duplication. Stratified random sampling was used as there existed a heterogeneous population, 2 homogenous population groups the Bagon Silan and Bago which lie within the boundary of the NNNP. This approach makes the sample reflect the population being studied. It provides better coverage of the population since there is control of the subgroups. There is no under representation of the population subgroup thereby reducing sampling error and the finding would be reflective of the population groups. To avoid duplication error, a list of households that had been awarded a CSC by the DENR under ISFP was obtained from the Provincial Management Office and verified with the above. The relevance of the sample was buttressed by consultation of the tribal leader to identify the indigenous people and their household.
Stratified random sampling ensures that the resulting sample will be distributed in the same way as the population (Bryman, 2016). Since the relevant information on population groups the process was quite feasible and more than one stratifying criterion could be used.
c) The interviews were done with the assistance of an interpreter and native speaker. Interpreter help respondents to understand the questions better in their own native language and thus they can also freely and correctly respond. Thus exclusion errors based on language were avoided.
d) Systematic and open ended questions
The use of systematic questions mean all respondents were asked the same questions and in the same order. This is good in that the same type of data will be collected and it makes analysis and comparison easier. If they do not have an idea, respondents can just answer what they know without being forced to lie based on set answers.
The use of open ended question also allow the interviewees to express themselves in their own words. As mentioned earlier the researcher can get more information on the feelings, attitudes and opinion regarding the situation in NNNP. Open ended questions reduce errors and bias as interviewees freely respond without any threats. Answers from open ended questions can be used for other future researches.
e) Quantitative data about cultivated parcel area, claimed parcel area etc was collected from each household. Quantitative data allows a broader study, involving huge number of subject and enhancing statistical generalisation of facts. Greater objectivity in the formulation of hypothesis ensures accuracy of results. The employment of prescribed procedures to ensure validity and reliability means the research can be replicated elsewhere and comparisons made across categories and over time. Personal bias can be avoided by researcher maintain distance from participants and employing strangers.
f) Statistical comparison using analysis of variance (ANOVA), Fischer exact test to compare groups for binary variables and t- test. Anova can be used to test for means of several populations.
g) The research also employed a mixed research approach which taps into the strength of both quantitative and qualitative methodology. The method also overcomes the shortfalls of the two approaches when implemented individually. It also works to confirm and explain findings through triangulation.
a) In depth household interviews
The challenges associated with interview is that respondents can give responses that they think they are social desirable instead of standing in the truth. Also because the interviews are in depth, they tend to be time consuming and affect results from the respondents at the end of each cycle when the interviewer is tired. Time is also consumed when transcribing, analysing and coding of the responses. There is also need to provide incentives to participants to motivate them and so avoid biased responses. Because of the time and incentive aspects, interviews are generally costly. It is not clear whether each member or a single member of the household was interviewed. If all were interviewed, it would further exacerbate the time and cost factors. However if one individual is interviewed, the danger would be on using his personal opinions to represent the opinions of the whole household. If the interviwer is not skilled and experienced the process is undermined.
b) Stratified random sample of 40 house hold was used in each Barangay was taken for a total of 80 household interviews. The list of household was obtained from the provincial management office and verified to avoid duplication. However the need to identify every member of the population being studied and classify it can be time consuming. It is difficult to find an exhaustive list of the entire population and classify every member (Investopedia, 19 August 2018) .
Also in real life the existence in multiple groups is unavoidable and if these are included in the sample it result in a biased sample. In this research it is also very difficult for a tribal leader identify all indigenous households. According to Bryman (2016) it is also uneconomical because the identification of the population for stratification is tedious since there is no listing in terms of strata
c) Interviews done with the assistance of an interpreter and native speaker
However the use of interpreters can be problematic if they misinterpret the question. This result in wrong responses being given based on wrong question. The other problem is on mistranslation which like interpretation can make the interviews very long. There problems are of interpreting non verbal, idiomatic communication and other native words which do not have an English equivalent. All these lead to loss of meaning and research information.
d) Systematic and open ended questions
The challenge with open ended question is that they can result in the interviews taking too long and the interviewer might not have written everything said. Interpretation, transcribing and analysis could be difficult especially from the 80 respondents. Open ended question result in inter-coding and intra-coding errors (Bryman, 2016)
e) Quantitative data about cultivated parcel area, claimed parcel area etc. The limitations with quantitative data usage and analysis lies in the collection of narrow and superficial data set. The use of standard question could cause structural bias as the data reflects the view of the researcher instead of participants. The results are often limited as they provide numerical description rather than a detailed narrative and will not reflect how people feel about or perceive a subject.
f) Statistical comparison using analysis of variance (ANOVA), Fischer exact test to compare groups for binary variables and t- test.
Anova test for means cannot be used to test which means is different from the others. Thus it needs support of other tests. Fischer’s test is conservative because its rejection rate is below the nominal level of significance. If there are extreme values the range is affected thereby making statistical analysis unrepresentative.