Data Analysis- lectures 1-3 Part 2

winniesmith2's version from 2017-10-16 12:34

Section 1

Question Answer
Examples of why experimental designs are not always feasible -Experiments often used to test a theory developed from an earlier survey or experiment. -Experiments normally created an artificial environment. -May not be able to manipulate the variables in real life environment. -Too costly.
What is the most commonly used tool for the collection of large dataQuestionnaires
Describe questionnaires -Good to study attitudes, values, beliefs and motives. -Reduced bias – questionnaires presented to everyone in the same way. -Open and closed questions (but often fewer open ended questions). -Used mostly for descriptive research -The study involves large sample -Can be self-administered or by a researcher -Less intrusive than telephone or face to face interviews -Mail or postal questionnaire is the most common form -Easy to enter data, & analyse (pre-code)
How can you pre-code questionnaires by using set closed responses. -Numbers can be entered or scanned into a pre-coded SPSS data sheet, rather than having to type the full word every time. -If the survey is internet based using a tool like Survey Monkey the data will be automatically dropped into a database for use in a software package like SPSS.
Designing the self completion questionnaire- guidance 1 1. -Easy to follow design – personal data at end -Do not cramp the presentation -Vertical or horizontal closed answers? -Identifying response sets in a Likert scale -Clear instructions about how to respond -Keep questions and answers together
Designing the self completion questionnaire- some guidance 2 2. -Long questionnaires = lower response rates & respondent fatigue (reduced internal validity) -Need to elicit accurate information, requires well worded questions and co-operation of respondents -Respondents need to understand the question, but also feel happy to ‘share the information’ desired by the researcher -Start and finish with ‘friendly’ questions. -Keep questions short – avoid 2 questions in 1 -Use simple language, avoid jargon
Designing the self completion questionnaire some guidance 3 3. -Avoid leading questions -Don’t presuppose information -Ensure questions produce variability -Test questionnaire in person with some trial respondents who meet your study’s inclusion criteria -Vary question format to avoid boredom -Include respondent id on every page
What can you use to catch attention logos,but need to be well designed and appropriate to target audience.
Advantages of self completion questionnaires -Cheaper to administer. -Quicker to administer - familiar to most people. -Absence of interviewer effect. -No interviewer variability. -Convenience for respondents.
Disadvantages of self completion questionnaires. -Cannot prompt or probe -Cannot ask too many questions not salient to respondent -Difficult to ask other kinds of questions -Questionnaire can be read as whole -Do not who answers/completes -Cannot collect additional data -Difficult to ask many questions. -Respondents can report ‘desirable answers’ rather than ‘real’ answers -Respondents may not take research seriously (e. g. religion in national censuses) -Not appropriate for some respondentslow literacy -Greater risk of missing data -Lower response ratehuge problem in quantitative research
What steps can you use to improve response rates -Good covering letter (if possible personalise it) -Return S.A.E. , with clear instructions -Follow up for non-response -Questionnaire - correct length -Must capture the respondent’s interest -Limit the number of open ended questions -Provide an incentive (e.g. a gift voucher)
Points for writing covering letters1. Friendly but short 2. Describe why study is being done 3.Tailor letter to audience 4.Mention incentives (if any) e.g. copy of results 5.Mention s.a.e. 6.Use deadline to get prompt response 7.Describe confidentiality/ anonymity policy 8.Name and phone no. (only corporate no. should be given not personal ones for safety reasons) for any questions
How can you improve response rates with follow ups and reminders. -Response rate often too low without them – non respondents often have different characteristics . -Should include 2nd copy of the questionnaire. -Follow up means you can’t guarantee anonymity only confidentiality. -In cover letter sound surprised at non-response.

Section 2

Question Answer
Describe diaries as a form of self-completion questionnaire 1 -Often called Researcher-Driven Diary. -Can be seen as self-completion questionnaire (if it is structured) but more likely it is an alternative method of data collection to observation (As it is free text). -Structured diary is accurate at tracking events as they occur. -Free text diary is used to record behaviour but coding this type of data carries huge difficulties.
If diary is to be used then you must; -Provide clear guidance -Be clear about the time periods -Provide a ‘model’ -Provide a checklist -Include designated times – start and finish
Advantages of diaries as method of data collection-Provide valid and reliable data -Good sequencing of behaviours -Can be used for sensitive behaviours
Disadvantages of diaries as a method of data collection-Tend to be expensive -People get fed up -People become less diligent -Problems of memory recall

Section 3

Question Answer
What influences the choice of design strategies (4 points to consider) 1. Is the experiment or cross sectional design strategy appropriate to your research/study? 2. Don’t rule out other possible strategies at this stage 3. Consider the Purpose of your Research/Study The Purpose of Study may help in selecting the strategy 4. Does the research question have a strong influence on the strategy selected?
Experimental study Measuring the effects of manipulating one variable on another variable
Cross sectional/survey Collecting large data using a standardised form from a group of people
Explain why you should not rule out other possible strategies at this early stage Different researchers prefer to use either experimental or cross sectional/survey as their way of working. A hybrid strategy may fall between these two camps and be best to use.
Explain a hybrid strategy -You can carry out a survey of small number cases but use a ‘large questionnaire’, (most likely to use a researcher) . -Similarly you can carry out an experiment for which the data is obtained by a survey, (most likely to use a self-administered questionnaire).
What do you need to consider when you are considering the purpose of your research/enquiryis your study an exploratory, descriptive or explanatory?
Exploratory To find out what happening. To seek new insights, to ask questions. To assess phenomena/ occurrences in a new light
DescriptiveTo accurately profile people, events, situations. Requires extensive knowledge of the situation, etc., to be researched so that appropriate data can be collected
Explanatory Seek an explanation of a situation or problem usually in a causal relationship.
Why may the purpose of the study help in selecting the strategybecause different methods are better for different studies. For example,Surveys are appropriate for descriptive studies, Experiments are appropriate for explanatory studies and Case studies are appropriate for exploratory studies - (not part of this module).
What types of research question lead to an experiment How, why (requires control over events? yes focus on current events? yes)
what types of research question lead to a survey who,what,where, how many, how much. (Requires control over events? no focus on current events? yes)

Section 4

Question Answer
Why do you need to consider if data are randomly missing? happens when - The respondent/subject actively decides not answer the question. This is a serious study design fault and can’t be overcome with more sophisticated analysis - hence need to reconsider study design.
What is pilot study 'a dummy run' tests measures you are going to use on a very small sample and helps to identify the feasibility of what you are going to do .
Why are pilot studies especially important in research with humans because people are very rarely predictable. As part of the pilot study – have your peers/ other scientists comment on your questionnaire.
What do you do if you think missing data is at random?less of a problem for design, enter data as missing using codes, but Do NOT fill in missing data with zeros. ‘999’ for missing is convention (Do not use 999 if it can potentially be a real value in the data) ‘.’ for system missing (for example first part of question not applicable so they don't answer second).
Steps to plan successful data analysis 1. Formulate a testable research question (your hypothesis) 2. Collect data appropriate to testing your hypotheses: +Consider the types of variable to need to answer your question: -Dependent Variable or Outcome Variable -Independent Variables or Predictor Variable +Scales of measurement-Continuous scales - Interval and Ratio measures -Categorical scales – Nominal and Ordinal measures 3. Decide on the type of analysis appropriate to test your hypothesis 4. Properly interpret and report your results