measures of movement Flashcards
what are the four measures of movement?
- gait (walking)
- biomechanics
- posture
- balance
describe objective measures for movement analysis
- lab- based movement biomechanics
- standardised, repeatable and precise
what are the examples of objective movements?
- 3D motion data
- trajectories
- joint angles
how are objective measures precise?
- exact joint angles
what are the disadvantages of scaling up?
- limited to small scale studies
- complex analyses of raw data
do we always need details of how they move?
- no, also need to know how much they move
what are the three ways of measuring physical activity?
- self- report
- pedometers
- research grade activity monitors
what are self report measure examples?
- Global PAQ
- GP- PAQ
what are the two main limitations of self report measures?
- open to bias and over estimation
- vague time periods
what are pedometers?
- small portable device that counts the number of steps a person takes
are pedometers simple?
- very simple operation
- requires a battery
what are the advantages of pedometers?
+ low cost
+ can monitor large samples at one time
what are the limitations of pedometers?
- only 1 dimension of data
what are research grade activity monitors?
- range from low- cost e.g., axivity to high cost e.g., actigraphy APDM
what is a axivity?
- data logger that includes MEMS 3- axis accelerometer
- categorises activity levels
what are the advantages of axivity?
+ simple functions
+ large sample
what is a actigraphy?
- monitors human- rest activity cycles using wrist- worn device
what is the advantage of actigraphy?
+ multiple functions
what units are acceleration and orientation?
- inertial measurement unit
how can we now track data? how does this work?
- can use smartphones and wearables to capture data on how much we move and when (24/7)
describe the prevalence of smartphones and wearables
- increasingly ubiquitous within population
what do smartphones and wearables allow us to understand?
- health
- lifestyle
- behaviour
what are some examples of major fitness trackers? what can you download?
- apple health kit
- google fit
- fitbit
- can download raw data
what are fitness apps designed to do and track?
- can automate the data capture
- long- term tracking of daily step count and other measures
how can we measure frequency?
- step count
- number of bouts
how can we measure intensity?
- cadence
- metabolic equivalent of task = 02/ energy expenditure during quiet setting
how do you measure time?
- periods of moderate- vigorous PA
- sedentary periods
how do you measure type?
- inclines/ stair climbing
- walking/ running
- sitting/ standing
what can the data cause?
- causes data overload
- alot of raw data collected
what is required to help us understand values?
- context is needed
what is the trial you can do with the data?
- randomised controlled trial
- compare placebo or healthy group with health condition and intervention
what test can be used with the data?
- pre- post test
- compare baseline prior to intervention
what other data can you combine data with?
- questionnaire data
- qualitative data
what does questionnaire data allow?
- allows grouping/ predictors of change
what does qualitative data give us?
- details on individual insights
what can we identify from data? and how?
- identify patterns using data- driven analyses e.g., machine learning
what ethical considerations are there when working with wearables/ phone data?
- privacy
- health inequality
- biases
what happens to data in relation to privacy
- data use by tech company behind device
- transfer of data for research
what questions can be asked regarding privacy when using phones/ wearables?
- where will the data be stored?
- will the data remain anonymous?
what is anonymised data ?
- all identifiable data is removed e.g., name, address, mobile number
- no way of tracking data back to an individual
what is pseudonymised data?
- same as anonymised but data has participant ID
- separate file contains ID and identifiable
can individual be tracked in pseudonymised data?
- people with access to file can track back to an individual
why may you think data is anonymous but it’s not?
- large datasets interlinked with many variables
- does removing the identifiable information make it fully anonymous?
what is the test for anonymity?
- motivated intruder test
what is the motivated intruder test?
- starts without any previous knowledge
- aims to identify an individual from an anonymised dataset by accessing resources
what resources are accessed in the motivated intruder test?
- internet
- libraries
- all public documents
what is motivated intruder test reasonably?
- reasonably competent
what does the motivated intruder test employ?
- employs investigative techniques
- including questioning people who may have additional knowledge of the individual
what is the person doing the test assumed to have?
- assumed to have no specialist knowledge such as computer hacking skills or any access to specialist equipment
what does the motivated intruder test not resort to?
- criminality such as burglary to gain access to data that is kept securely
how do you add extra layers of privacy?
- keeping detailed variables
- always a chance an adversary will be able to de- anonymise data
what should you remove to add an extra layer of privacy?
- remove all signals in a dataset to anonymise
- can make some analyses pointless
what is the solution of anonymity?
- randomisation
- differential privacy
what can you insert to help anonymity?
- insert random noise into the information made available
- done correctly, meaningful answers can still be retrieved
what is an example of randomisation?
- participants told to flip a coin before answering the question
Heads= give real answer, the truth
Tails= answer randomly e.g., flip another coin to determine yes/ no
what is the health inequality relating to smartphones/ wearables?
- exclusion if can’t afford the device
what are the other health inequalities?
- impact on access to healthcare for those with and without data
- accessibility/ usability of the technology (age exclusion)
what are the biases in terms of expenses of devices?
- more expensive devices = better data quality?
what is the biases with step count ?
- might only work with healthy gait
what are the biases with Ai models?
- trained on representative data
what are the potential issues with the activity trackers?
- quality/ validation
- how accurate are trackers?
what are activity trackers designed for?
- designed for healthy population
how can slow gait or asymmetrical gait cause issues?
- slow gait causes issues e.g., heel strikes and toe offs less pronounced
- asymmetrical gait, other pattern changes > affects algorithms
describe incentives for physical activity case study
- access to the app’s 6000 users, anonymised data
- daily step count; 6 months since registration but also 3 months prior to registration baseline measure
what are the three factors of data cleansing and formatting?
- outliers
- data variability
- missing variable
what are outliers in data cleansing and formatting?
- are there anomalous data points/ users?
- excessive step counts?
- remove data point/ user
what is an example of an outlier?
- 3 SD above the mean
what is data variability in data cleansing and formatting?
- does the mean step- count of an individual vary widely each month/ week
- remove user
what is an example of data variability?
- top- percentile of variance
what is a missing variable in data cleansing and formatting?
- does an individual have many days of missing values?
- remove user
what is an example of missing variables?
- > xx%
- n days missing
what is used as reference data? what does it define?
- registration date
- defines monthly interval
what does objective data variable include? describe variable
- daily step count
Dependent variable = change in step
what does subjective/ questionnaire variable include? describe variable
- questionnaire
Independent variable = age, gender
how do you work out simple regression? - describe different letters
y = B0 + B1x
y= DV; x= IV ; B0= intercept
- relationship between x and y described by B
how do you work out multiple regression? what relationship is shown?
y = B0 + B1x1 + B2X2 + BnXn
- what is the relationship between y and x1, x2, xn
- defined by B
what regression types are used for continuous/ dichotomous class?
- continuous= regression
- dichotomous= logistic regression
what is B in logistic regression?
- increase in ‘log odds’ of the outcome (DV) per unit increase in IV
what do we take from B in regression and why?
- complicated to understand so we take the exp (B) to get the odds- ratio (OR) relative to reference category
what does OR= 1 mean?
- IV doesn’t affect outcome
what does OR> 1 mean?
- category associated with higher odds of outcome relative to reference
what does OR < 1 mean?
- category associated with lower odds of outcome relative to reference
