Biology simple sampling

Classified in Mathematics

Written at on English with a size of 25.21 KB.

 

Quantitative Data Collection:

Types of producing quantitative data are types of tests: language tests, psychological tests, etc./ Behavioural frequency: Measurement of phenomena in an objective way (certain types of behaviour frequency in a certain period of time/ Survey by using a questionnaire (Besides tests, are the most used in applied linguistics)

Sampling in Quantitative Research: The decision must be taken early in the overll planning process. Common questions when making research, since participant sample could determine the success of a study: How large should my sample be? Who shall my sample consist of?

There are three main concepts in quantitative samplingSample: Group of participants whom the researcher actually examines in an empirical investigation. For instance, 4 members of the LGBT community. Population: Group of participants whom the study is about. For instance: LGBT community. Notion of representativeness.

  • Target population of a study consists of all the people to whom the survey’s finding are to be applied or generalised.
  • A good sample almost equals the target population in its most important characteristics: age, gender, ethnicity, social class, etc.
  • Sample is a subset of population that is representative for the whole population
  • Milroy and Gordon: Representativeness is crucial. The strength of the conclusions we can draw from results obtained from a selected group depends on how accurately the particular sample represents a large population.
  • Even though, census uses inclusion of every member of a population in a study, investigating the whole population is not necessary and could imply a waste of resources.
  • Appropriate sampling procedures: Saves time, resources and effort but obtains accurate results.

Sampling Procedures.

  • Two types of sampling strategies:
    • Probability Sampling: Complex and expensive procedures. (Scientific procedures use)
      • Random Sampling: Selects members on a completely random basis. Based entirely on probability and chance: Minimises the effects of strange or subjective factors. The sample should contain subjects with similar characteristics to the population as a whole. It is rarely fully achieved, but they are more representative.
      • Stratified Random Sampling: Combination of random samples with rational grouping (Randomization and categorization). Population is divided into strata, and random samples of each strata.
      • Systematic Sampling: In cases of random selection, anonymous survey can be hard to achieve, in this case, systematic sampling can be used. This sampling selects very Xth member of the target group.
      • Cluster Sampling: Way of making random sampling more practical: Randomly select some larger groupings or units population and then examine all subjects in those selected units. (For instance, pick a school and analyse the students).
  • In applied linguistics it is not feasible to get perfect representativeness.
  • Non-probabilistic Sampling: Try to achieve representative samples with ordinary resources. This is a common technique in applied linguistics and in qualitative data. There are two main non-probabilistic strategies:
    • Quota sampling and dimensional sampling: Start with a sampling frame and then determine the main proportions of the subgroups defined by the parameters included in the sampling frame. The actual sample then, is selected to reflect these proportions. No random selection within the groups, but the investigator selects participants. (For instance: 50% are monolingual families and 50% are bilingual families. 150 members of each group will be selected not randomly.)
    • Snowball sampling: Involves chain reaction whereby the researcher identifies a few people who meet the criteria of the particular study and asks the participants to identify appropriate members of the population for the study. This is apt for groups whose membership is not identifiable or when access to a group is difficult.
    • Convenience or opportunity sampling: Most common in L2 research. One important criterion of sample selection is the convenience of researcher: Members are selected if they meet certain practical criteria, like geographical proximity, time availability. Besides accessibility, participants have to possess certain characteristics related to the purpose of the study (Partially purposeful)
  • Non-probabilistic sampling is rarely generalizable.
  • Kemper: Sampling issues are practical: It is in sampling where theory meets the hard realities of time and resources: they force pragmatic choices.
  • Limitations of non propapilistic samples.
  • Characteristics of the particular sample shared with target population
  • Careful about the claims relevance of our finding

How large should the sample be?

  • There are no hard or fast rules in setting the optimal sample size. Nonetheless the research should consider guidelines:
    • Rules of Thumb: In survey research it should be from 1% to 10% of a minimum of 1000 participants could represent the right sample. The more scientific, the less participants are recommended. In certain types of quantitative methods some estimates of sample have been made:
      • Correlational research: At least 30 participants.
      • Comparative and Experimental Procedures: At least 15 participants.
      • Factor Analytic and Other Procedures: At least 100 participants.
    • Statistical Consideration: The quantitative research should have a normal distribution.
      • Hatch and Lazarton: To achieve normal distribution the sample should have at least 30 persons and if not, this could be compensated by the usage of special procedures.
    • Sample Composition: Consider if there are any distinct subgroups within the sample that may be expected to behave differently from others. In case it is like this, the sample size should be set so that the minimum size applies to the smallest subgroup of the sample.
    • Safety Margin: When setting the final sample size it is necessary to set a margin that takes in count unforeseen or unplanned circumstances (Dropout some phases of the project or subgroup that needs to be treated separately)
    • Reverse Approach: Statistical significance depends on the sample size. I s the sample big enough to reach statistical siginifcance?. Determine expected magnitude or power of the expected results and then determine if one sample size is necessary to detect if one effect actually exists in population.

The problem of respondent self-selection.

  • Participant self-selection can represent a threat for the validity of the investigation. When the sample is formed in part by participants who decided to participate and not only the function of some systematic selection process. Participant self-selection can be exemplified when:
    • Researchers invite volunteers to take part of the study. (Offering money or treatsjj)
    • The design allows a big number of dropouts (mortality), in this case, participants self-select themselves out of the sample.
    • Participants are free to choose whether they participate in a study or not (Postal questionnaire surveys).
  • Participant self-selection is almost inevitable, but in cases like the ones mentioned above, could result in a sample that is not similar to the target population. For instance, volunteers may be different from non-volunteers in their attitudes, motivations, characteristics, etc. So the sample can lose representativeness (generalizability).
  • Brown: If the respondents of a questionnaire have certain characteristics and others have other characteristics, then the results cannot be applied to both groups.

Questionnaire Surveys. Main issues of surveys:  how to sample participants/ How to design and administer the research tool

  • Questionnaires are very popular in social sciences: its popularity is given by their easy construction, versatility and capacity of gathering lots of information quickly in a form that is readily processible.  Questionnaires and tests of proficiency are the most used methods in language research (Applied linguistics). In applied linguistics, though, it is very common to come across questionnaires that fail, because they can lack of reliability and validity.

Questionnaire Theory.

Questionnaires can be understood in two broad senses: Interview schedules/guides /  Self administered pencil-and-paper questionnaires.

Brown: Any written instrument that presents respondents a series of questions or statements to which they are to react writing the answer or selecting from one existing answer.

  • Questionnaires can get three types of data about the respondents:
    • Factual Questions: Certain facts about respondents like age, gender, socioeconomic status or amount of time living in a L2 community.
    • Behavioural Questions: Obtain the current life of respondents and what they have done in the past (Habits or lifestyle)
    • Attitudinal Questions: What people think, their opinion and interests.
  • Even though they are similar to written tests, they are different because:
    • Written tests evaluate the respondent’s underlying competence (like proficiency)
    • Questionnaires are non-evaluative instruments: do not have good or bad answers.

Multi-Item Scales. How items of a questionnaire are worder have an important implication in the responses, minor differences in the formulation of questions produce radically different levels of agreement or disagreement, for instance, changing a vocabulary item by another with the same logical meaning can have an impact on responses.

Rensis Likert: Wording problem in questions can be solved with multi-item scales.

Multi-item Scales: Cluster of several differently worded items that focus on the same target. The item scores for similar questions are summed resulting in a total scale score. The item’s interpretation will be given by the summation of the item’s scores.

To maximize the stable component that items share and reduce influences related to individual items, 4-10 items should address the same area with slightly different aspects of it. Specific character of questionnaires questions make them very suitable for quantitative and statistical analysis, even though there can be open-ended questions.

Open ended questions: Tend to be superficial with the topic investigated.

Closed ended questions: Permit codification of answers with certain scores.

    • Formats of closed-ended questions:
      • Likert Scales: Characteristic statement, and respondents are asked to indicate the extent in which they agree or disagree.
      • Semantic Differential Scales: Elicits graduated responses as the Likert Scale. This technique is useful because researchers avoid writing statements. Respondents are asked to mark their answers in a continuum between two opposite adjectives. The importance of this technique is that it is easier to avoid superficial answers by alternating the sides of positive or negative poles.
      • Numerical Rating Scales: Assignation of one of several numbers in a scale to describe a feature of the target (Como cantidad de estrellas pa una pelicula). The positive aspect of this technique can refer to a wide range of adjectives (excellent, good or awful).Other closed-ended item types: Depending on the type of questionnaire, age of participants and other characteristics.
      • True-false items: Used in cases where polarized yes-no answers are considered reliable, like in the case of children who are unable to provide more elaborated ratings. The problem with this method is that it can simplify things too much, resulting in reduced or distorted data.
      • Multiple choice items: Popular in applied linguistics and to test L2 proficiency. Regularly used to ask about personal information of participants.
      • Rank-order items: Responds to human nature to rank information and concepts. They contain a list and participants are asked to order the items by assigning a number according to their preferences.
    • Open-ended questions:  Questions where are no response options, but rather there are lines for respondents to fill in. A benefit related to this type of questions is that it permits greater freedom of expression, so richer quantitative data. Sometimes they are necessary because researchers are unable to know the possible answers and cannot provide prepared responses. According to the author, this type of questions can work if they are not completely open, but have some guidance as in:
      • Specific Open Questions: Specific answers are looked like facts or habits about the participants.
      • Sentence Completion: Incomplete beginning of a sentence is provided and the respondent completes it according to his or her preferences.
      • Short-answer Questions: Designed for answers to be succint and short: more than a phrase but less than a paragraph.

Rules about item wording. : Item design is not a 100% scientific activity because in order to write good items one also needs a certain amount of creativity and lots of common sense. Even though there is absence of hard and fast rules, some advices can be specified: Aim for short and simple items/ Use simple and natural language/ Avoid ambiguous or loaded words and sentences: Avoid negative constructions/ Avoid double-barrelled questions: Those that ask two or more questions in one while expecting a single answers. Avoid items that are likely to be answered the same way by everybody/ Include both positively and negatively worded items

The format of the questionnaire.main Parts: Questionnaires have a standard component structure. Title/ General Introduction/ Specific Instructions / Questionnaire Items: Constitution of the main body of the questionnaire. They have to be separated from instructions/  Additional Information: Include contact information Final “Thank you” Length:  Most researchers agree that questionnaires should have a length of 4-6 pages.

  • Layout: Good design of the questionnaire can act as a motivation to respondents to produce reliable and valid data (They will think that the survey is professional and serious):
    • Booklet format: Has to look short and has to make it easy to read and to turn pages.
    • Appropriate density: We must not make the pages look crowded: reduce margins, certain types of fonts, etc.)
    • Sequence marking: Mark each section of the questionnaire with markers: roman numbers, arabic numbers.
  • Item Sequence: The order of the items is significant because the content of a question can have an impact on the interpretation and the response:
    • Mixing up scales: Different scales need to be mixed up to create a sense of variety and to prevent repeating answers.
    • Opening questions: They need to be interesting, simple and focused on salient aspects. It has to be mild or neutral.
    • Factual questions: Oppenheim claims that factual or personal questions are best left at the end of the questionnaire. Issues like age, level of education, etc are considered as factual questions, and need to be answered at the end, because if they are placed at the beginning they can create resistance in the respondents.
    • Open-ended questions at the end: If not, the other items may be affected by the potential negative consequences of the open-ended questions, like the required work that they imply.

Quasi-experimental Design.

  • Similar to true experiments in every respect except that they do not use random assignment to create comparisons to see the treatment-caused changes.
  • Heinsman and Shadish, when comparing quasi experimental and experimental designs found that if the two research methods were equally well designed, they can have comparable results, but it's not easy.
  • To improve this, it is necessary to improve the design of quasi-experimental designs by:
    • Avoiding situations of self-selection
    • Minimize pre-test differences between the treatment and the control groups as much as possible. For this goal you can:
      • Match participants of the treatment and control groups. In quasi-experimental design this is not feasible.
      • Using analysis of covariance (ANCOVA): STATISTICAL METHODS FOR ADJUSTING THE POST-TEST SCORES FOR ANY-PRE-TEST DIFFERENCES.
  • Quasi experimental design is more susceptible to threats to validity, but just because a threat is possible it does not mean that it is plausible. )Decrease plausibility is possible by applying the above techniques). Quasi experimental study that is well designed and executed can have scientifically credible results.

Strengths and Weaknesses of Experimental and Quasi-experimental Designs. Main strength of experimental design is that it is the best method to establish cause-effect relationships and evaluating educational innovations. Nonetheless the price of this can be high, he exposes that in order to control the variables it is necessary to apply artificial frameworks with reduce external validity (Clarke) another weakness of experimental design is the Hawthorne Effect: The outcome of the study is not caused by treatment variable, but by the fact that a treatment has been applied, regardless its nature. In quasi experimental studies, the main benefit is that it takes place in authentic learning environments with genuine class groups, but this benefit  is also associated to threats because of the inequality of the initial treatment and control groups. The selection bias in this case involves differences that are not due to the treatment but artefacts of the differences in pre-existing characteristics of groups being compared. In this case, the list of possible confounding variables is sol large that achieving satisfactory control is a challenge. In educational contexts it is possible to have a whole range of potential independent variables to test, so experimental and quasi experimental design is not feasible


Strengths and Weaknesses of Questionnaires.

  • The main benefit of questionnaires is the efficiency of researcher time and effort and financial resources: One can collect much information in less than an hour. Processing the data can also be fast. Questionnaires are also very versatile (Used buy many people in many situations and regarding many topics. Respondents do not mind about filling questionnaires,
  • Limitations of questionnaires include; the easy production of unreliable and invalid data by means of an ill-constructed questionnaire. Gilham exposes that no single method has been so abused. The weakest aspect of questionnaires is the fact that items need to be sufficiently simple and straightforward to be understood by everyone. Its simplicity can signify superficial data, the short time of its application implies that the investigation is not very deep.

Experimental and quasi-experimental studies. Experimental study is another type of quantitative data collection, it is the most scientific method because it can establish unambiguous cause-effect relationships. It is hard to establish cause-effect relationships because in real life nothing happens in isolation and there are many interferences between the related factors

Experimental design: Take a group of learners and do something special with them, while measuring the effects of this. Then compare their results with data obtained from another group that is similar in every respect to the first group except for the fact that it did not receive the special treatment. So the discrepancy of results between groups can be attributed to the difference of treatment between them.A typical experimental design would be an intervention study of two groups: Treatment or experimental group: Receives the treatment.

Control group: Provide a baseline for the comparison

Collecting Quantitative Data Via Internet

  • Through the growing use of Internet, researchers have tried to collect data by means of web-based procedures.
  • Internet-Based experiments have certain benefits:
    • Reduced Costs: It is not more expensive than traditional research.
    • Convenience of Administration: It does not require administration of tests (Self-running)
      • Automatic Coding: Codification and recording of the answers is automatic through CGi script.
      • High Level of Anonymity: Enhances levels of honesty.
      • International Access: Larger and diverse samples worldwide, cool in Cross- cultural research.
      • Access to Specialized Populations: Access to small or specialized populations which in real life could be difficult to reach.
  • Technical issues associated to internet based research include: Differences in computers, systems, browsers and monitors in internet users. The questionnaire may not be available in the same format to as wide a population.
  • Sampling issues in Internet based research include: that it is not possible to apply systematic strategy of sampling.  There is a lack of control over who will eventually participate in the study. The actual sample may be much more heterogenous than in traditional research, because it only consists of self. Selected participants, this means that it would be difficult to generalize findings. Birnbaum postulates a partial solution for the sample problem: Analysing the research questions separately within each substratum of sample (age, sex, etc), then if  results are similar, it can mean the external validity of results. The second solution is the comparison of subsample, of traditional survey and internet-based survey, creating new findings. Also by combined studies, the sample can be approached in traditional way and ask them to complete a survey or experiment online.

Entradas relacionadas: