Presentation on theme: "APA6101 INTRODUCTION. Course goals Understand the basics of experimental design as it pertains to research in kinesiology. Be aware of issues surrounding."— Presentation transcript:
Course goals Understand the basics of experimental design as it pertains to research in kinesiology. Be aware of issues surrounding privacy and ethics when doing research on human populations. Know how to set up a research data base using SPSS. Appreciate the basic principles underlying statistical analysis. Develop an ability to decide on the most appropriate statistical analysis to use for various research questions. Know when to call a professional statistician. Be able to implement simple analyses in SPSS. Know how to write a coherent statistical methods section and a results section for a manuscript.
Evaluation Mid-term Exam. 20% In –class presentation. 10% Assignments. 2, for a total of 10% Report and data analysis. 35% Final exam. 25%
1.Setting up the study 2.Statistical principles. Comparing two samples 3.The linear model (Regression) 4.The General Linear Model (comparing several samples) 5.Binary outcomes: odds ratios and logistic regression 6.Contingency tables 7.Repeated measures 8.Other topics (to be discussed) 9.When the normality assumption fails: non-parametrics and the bootstrap 10.Systematic reviews and meta-analysis
Agenda for today Introduction. Goals and expectations. Study design and randomization. ◦Primary and secondary research questions. ◦Recruitment, privacy and ethics. ◦How many subjects do I need? (More on this later) Setting up a data base in SPSS. Simple summary statistics using SPSS. The histogram
Study design Primary research question ◦The design and sample size calculations are geared towards this. Secondary research question ◦Other questions that the data can answer.
Example Kinematic and electromyographic comparisons between chin-ups and lat-pull down exercises. Kenji Doma, Glen B. Deakin, Kevin Ness Sports Biomechanics, 2013. Vol. 12, No.3, 302-313
Abstract The purpose of this study was to compare kinematics and muscle activity between chin-ups and lat-pull down exercises and between muscle groups during the two exercises. Normalized electromyography (EMG) of biceps brachii (BB), triceps brachii (TB), pectoralis major (PM), latissimus dorsi (LD), rectus abdominus (RA), and erector spinae (ES) and kinematics of back, shoulder, and seventh cervical vertebrae (C7) was analysed during chin-ups and lat-pull down exercises. Normalized EMG of BB and ES and kinematics of shoulder and C7 for chin-ups were greater than lat-pull down exercises during the concentric phase (p < 0.05). For the eccentric phase, RA during lat-pull down exercises was greater than chin-ups and the kinematics of C7 during chin-ups was greater than lat-pull down exercises (p < 0.05). For chin-ups, BB, LD, and ES were greater than PM during the concentric phase, whereas BB and LD were greater than TB, and LD was greater than RA during the eccentric phase (p < 0.05). For lat-pull down exercise, BB and LD were greater than PM, TB, and ES during the concentric phase, whereas LD was greater than PM, TB, and BB during the eccentric phase (p < 0.05). Subsequently, chin-ups appears to be a more functional exercise.
Testable question The purpose of this study was to compare kinematics and muscle activity between chin-ups and lat-pull down exercises and between muscle groups during the two exercises.
Study types Descriptive or population study ◦What is the prevalence of ….? ◦How many …. ? Comparative ◦Intervention study (compare treatment and control) ◦Case-control. Work backwards to describe differences in cases and controls ◦Comparing several groups Studies can also be characterized by the design type ◦Observational ◦Trial (experiment, clinical trial)
Chin-up paper Primary research question ◦Compare the EMG responses of the selected muscle groups between the chin-ups and lat-pull down exercises and to determine whether the EMG responses were influenced by the kinematics of the exercises. Testable hypotheses ◦The muscle activity of the primary movers and core muscle groups and the displacements of the selected body segments would be greater for chin-ups than for lat-pull down exercises. ◦The primary movers (BB and LD) would produce a higher level of muscle activity over the non-primary movers during chin-ups and lat-pull down exercises.
Comparative study Comparing chin-ups to lat-pulls Comparing one group of muscles (major movers) to a different group (non major movers) across exercise type Each participant is his own control ◦One might have given half the group one exercise, half the other, and compared group outcomes ◦In this case, each participant does both exercises. Another possible question: Does the relative importance of major to non-major movers differ according to exercise type? (i.e. is there a muscle group x exercise type interaction?). They don’t seem to have asked this question.
Design and protocol in the chin-ups study Subjects are selected according to specified inclusion criteria ◦healthy males with at least 12 months resistance training. ◦Informed consent was obtained prior to the trial. ◦All procedures in this study were approved by the James Cook University Human Research Ethics Committee. Inclusion criteria and ethics considerations are stated in the Methods section of a paper.
Selection criteria RANDOMIZATION, SAMPLE SIZE, PRIVACY AND ETHICS
Randomization Participants should be selected randomly from the population under study. In a comparative, intervention study, assign the intervention (or control) to participants randomly. ◦this is known as an RCT (randomized control trial) Where individuals are their own controls, do the intervention and control in random order. That was done in the chin-ups study.
How to randomize Use a mechanical or computer assisted tool ◦Dice ◦random number generator (pseudo-random numbers) ◦Block randomization (to get equal numbers) A person is not a random number generator In an intervention trial, ◦pre-allocate individuals to treatment and control group. ◦Keep the allocation secret (sealed envelopes). ◦The randomization should not be done by the investigators, but by a neutral third party. Details of the randomization scheme could be included in the Methods section.
Randomization and recruitment Hospital setting 1.Nurse approaches patient meeting inclusion criteria to see if they are willing to hear about the study? 1.A medical study is being conducted in this unit. Would you be interested in hearing about it from the research co-ordinator? 2.If yes, The RC approaches the patient, explains the study and asks if the patient would like to participate. (It’s called informed consent) 3.If the patient agrees, at that point, the patient is randomized to intervention or control. Consent cannot be conditional on participating in one or other arm of the study. 1.I’ll join if I can be in the control group. NOT.
Table 1 (report recruitment history) Analysed Outcome Consented Approached Eligible 840 799 interested 400 Yes 200 Completed 190 complete information 10 some missing data 2 died Terminal information available 198 dropped out 50 available for follow up 148 unavailable to follow-up 399 No41 declined
Bias Recruitment bias ◦Volunteers are different from refusers ◦How to avoid this: retrospective studies based solely on patient chart information can generally be used without consent. ◦However we often just have to live with this. ◦Logistic difficulties restrict the degree of randomization possible and the degree of selection. ◦This limits the generalizability of the results ◦Meta-analysis allows pooling of results over a number of small studies. ◦Cross-sectional studies favour longevity Retention bias ◦Completers are different from non-completers
Sample size Calculate the number of participants you need to achieve the accuracy you want ◦Some prior knowledge is necessary of how variable the data will be. ◦You need to determine and effect size of interest (the smallest difference that you care about). Sample size calculations can get difficult and messy. ◦Contact a statistician. ◦We will see some simple cases later in the course.
Ethics Navigating the Ethics of Human Research In 2001, Canada's three federal research agencies, CIHR,NSERC and SSHRC,CIHRNSERCSSHRC jointly created the Interagency Advisory Panel on Research Ethics (PRE or the Panel) as part of a collaborative effort to promote the ethical conduct of research involving human participants. The Panel develops, interprets and implements the Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans (TCPS). See: Panel on Research EthicsPanel on Research Ethics These guidelines apply to research carried out under these granting bodies.
Highlights A research centre (hospital, university, or other) will have an ethics review board (REB) to review and approve proposed research. The following criteria may be relevant: ◦Safety and dignity of the subjects ◦Scientific value ◦Privacy, use of subject information Work with the REB. These people will keep you out of trouble!
Core values – safety and dignity Research is a value The dignity of the human individual must be respected ◦informed consent ◦Youth? the cognitively impaired? ◦participation in a study must not entail unreasonable risks. Some risk may be unavoidable. ◦Participants must not receive a treatment known to be inferior. ◦This may lead to consideration of sequential trials ◦Each individual continues to be treated as an individual in light of their needs. Proper treatment is not a subordinate value to research. ◦Care must be given not to arbitrarily exclude or include certain groups. ◦pharmaceutical trials performed on prisoners? ◦Medical trials limited to healthy, young men?
Scholarly criteria Depend on the institution and the discipline. An REB may take issue with the following: The study has been done before, and is thus futile. ◦Do a systematic review. Convince yourself that you have something new to say. The study looks incompetent when peer reviewed. ◦It could happen … just sayin’ ◦Nothing is said about how the data will be analysed, giving the impression that you don’t know what you are doing. The sample size is too small to give assurance that an effect of interest can be detected. ◦Consult a statistician FIRST.
Privacy Identifiable information should not be published Care should be taken while collecting and storing data to anonymize the data as much as possible ◦If a given person does not need to see particulars, then that person should not see particulars. Data collected for one study cannot be used in another study without obtaining information from Ethics ◦people consented to Study A and cannot be assumed to consent to Study B. Information used for the purposes of quality control or employee training are not subject to REB regulations. In general, if it’s going to be published, go through the REB.
An REB submission Typically, an REB submission looks like the first part of a research paper (or a grant application): ◦Statement of the research question ◦Summary of the current state of research ◦Why existing research leaves some questions unanswered ◦How the proposed research will contribute something new ◦The protocol and methods to be used, including ◦Inclusion criteria ◦Randomization and other design features ◦Proposed method of analysis ◦Secondary research questions that may be addressed through this project ◦A bibliography
Collecting data Collected from charts Interview data Electronically gathered Interpretation of images, videos Coding experience Self-report
My recommendations 1.Create a detailed data entry form to record every piece of information that will be collected. 1.Do not attempt to populate a data base directly from primary sources. 2.Set up the data base to reflect the data entry form. 3.Enter the data. 4.Check the data. 5.Record the raw data. SPSS can calculate derived variables. 6.Maintain an archived copy of the raw data (without derived variables)
SPSS A flat file data vault ◦Very good. Better than Excel A fleet of data analytic modules ◦Reliable but conservative A syntax language (for data manipulation and other uses) ◦Clunky and horrible Alternate strategy! Use R!
Two screens Data view ◦This is where you can enter and admire your data Variable view ◦Where variables are defined
Variable types in Excel Numeric ◦Scale. Actual number. ◦Ordinal. A rank ◦Categorical. Use numeric code. Annotate with descriptive labels ◦Dates. Number of elapsed seconds since midnight, October 10, 1584. String ◦Names ◦Identifiers
For Values (Gender), I specified 0=Male, 1=Female Always start labels from 0 (my recommendation)
My recommendations An SPSS variable name can contain some special characters. I stick to alphanumeric and “.” Note: names begin with letters ◦Q1.1 works ◦QuestionOne ◦questionone. This is the same as QuestionOne. Use camel case for legibility. Name consistently. Use the label field creatively. You can put a very long descriptor in this field. All categorical data that will be analysed should be entered as numbers with descriptors (values) added. Save the String format for comments and possibly subject identifiers