Module 1 Part 1 Research Design Flashcards
Different levels of analysis
Univariate
Bivariate
Multivariable
Multivariate
Univariate analysis
Data Summary - single variable
Bivariate analysis
Two variables
Multivariable analysis
Multiple independent variables
Practical steps for data analysis
- Data needs to be prepared for computer entry
- Data needs to be entered into a computer, stored and managed
- Data needs to be summarised and analysed to inform the research question(s) and draw conclusions for the study
Data management stages
- Coding (pre coding or coding after collection)
- Database or spreadsheet design
- Data entry and/or transfer
- Data cleaning - invalid response codes, missing data, duplicate data
- Security and subject confidentiality
Initial checking of questionnaires
Invalid responses
Check for missing data
Illegible writing
Data coding
Coding converts collected information into a suitable format for easy data entry and statistical analysis
Coding mostly converts text responses to numerical responses
Cording manual defines the coding system used
Coding manual
Integral part of data collection and management
Especially important in large studies involving different people
What should a coding manual contain
Definition of each variable
Range of expected responses
Format and instructions for recording data
Instructions for removing uncertainty
Recording of decisions regarding coding
Data entry
Can be into a spreadsheet, database or statistical software
1 data cell = 1 value
Statistical software
Usually looks like a spreadsheet
Has single row for each participant
Each colum represents one variable
Some packages have rules about length and style of column names
Dataset
Name given to a database in statistical software
Vulnerable to human error at several points - coding (researcher), data entry (research or data entry person), data consistency (respondent)
Data cleaning
Screening process to minimise occurrence of mistakes and corresponding impact on results
Accurate data - building blocks of analysis
Errors threaten validity of measures and impact on data analysis and results
Checking for coding errors and outliers
Verify data entry
Check data consistently
Strategies for increasing accuracy of data entry
Double entry - 2 independent enteries of the original data, discrepancies are checked and corrected
Random checking - subset of records are checked against original data
Create data entity template/form that does not accept data that doesn’t fit acceptable values
Consistency checking
Looking for inconsistent responses across questions in data file
Inconsistencies can be due to coding, typographical or respondent error
VERY IMPORTANT
Should you log all changes during data cleaning
Yes
Need to keep a record of all changes made during data cleaning
Record all coding decisions so that they can be used to refine coding manual
A surveyor was conducted of all nurses employed at a particular hospital. Among other questions, the questionnaire asked about the grade of the nurse and whether he/she was satisfied with their career prospects. What type of study design is this?
Quasi experimental design
Cross sectional study
Cohort study
Matched case control study
Cross sectional - takes snapshot of the characteristics of the nursing staff and their attitudes at a particular point in time
Which of the following is the most appropriate way to randomise participants in a randomised controlled trial
1. Researchers/investigators selects the treatment arm (intervention or control) using a random number generator and clinical staff are not told of allocation
2. The clinician asks the participant whether they would like to be in the intervention or control arm of study
3. Researchers/investigators selects the treatment arm (in verve riot or control) using a random number generator and tells the clinical staff which treatment to give
4. The clinical chooses which treatment the participant will receive
Researchers/investigators selects the treatment arm (intervention or control) using a random number generator and clinical staff are not told of allocation
A new triage system has been set up at the local Accident and Emergency Unit. To evaluate the new system, the waiting times of patients were measured for 6 months and compared with the waiting times as a comparable nearby hospital. What type of study design is this
1. Cross sectional study
2. Cohort study
3. Matched case control study
4. Quasi experimental design
Quasi experimental design - some features of experimental design such as testing an intervention and comparing it to a control but lacks randomised patients
20 high school students from school a and 20 high school students from school b were interviewed about their use of sunscreen. Each student was interviewed at the start of the school year and also at the end of the semester about a sun awareness eductional campaign
The observation units in this study are ___. There are ___ observational units and ____ observations
Students
40
80
