Week 8

Question 1

Lab based movement biomechanics analysis measures?

Answer 1

3D motion data
Trajectories
Joint angles

Question 2

Physical activity measures? Examples (if necessary)?

Answer 2

Self report measures eg global/GP PAQ

Pedometers

Research grade activity monitors eg APDM opal, Actigraph GT9

Question 3

Measures of movement ?

Answer 3

Gait (walking)
Biomechanics
Posture
Balance
Activity

Question 4

– of Self report measures?

Answer 4

Open to bias and over estimation
Vague time periods

Question 5

+/- of pedometers?

Answer 5

Low cost

Large sample activity monitoring

Simple operation (battery)

Only 1 dimension of data = step count

Question 6

+/- of Research grade activity monitors?

Answer 6

Some low cost, with simple functions for larger samples

Some are too high of a cost with multiple functions

Question 7

What are the inertial measurement units (IMU)

Answer 7

Acceleration and orientation

Question 8

+ of Phones and wearables in terms of data?

Answer 8

Capture data on how and when we move - 24-7

Increasingly seen everywhere within the population.

Opportunity to understand, health, lifestyle, behaviour.

All major fitness trackers allow raw data download eg Apple Health Kit, Google Fit, Fitbit

Long-term tracking of daily step count and other
measures, e.g. 2-years

Apps can automate the data capture

Question 9

What can we measure? With examples?

Answer 9

Frequency - step counts, no of bouts

Intensity - cadence, MET

Time - Periods of moderate/vigorous, sedentary periods

Type - Inclines, walking, running, sitting, standing

Question 10

What is a randomised control trial?

Answer 10

Comparing a placebo or healthy group with group with health condition and intervention

Question 11

Summarise privacy ethical considerations with
wearables/phone data?

Answer 11

• Data being used by the company that created device
• Transfer data for research
• Where is it stored
• Will it remain anonymous?

Question 12

Anonymised and pseudonymised data?

Answer 12

Anonymised Data - All identifiable data is removed (e.g. name, address, mobile number). No way of tracking data back to an individual

Pseudonymised Data - Same as above but data has ID.
Separate file contains ID and identifiable variables. People with access to file can track back to an individual

Question 13

How to test if datasets are actually anonymous ?

Answer 13

Motivated intruder test

Aims to identify individual with no knowledge

Assuming there is no access to resources, no investigate techniques, no specialist knowledge and no resorting to criminal techniques such as burglary.

Question 14

Summarise health inequality ethical considerations with
wearables/phone data?

Answer 14

Excluded if device cant be afforded

Impact on access to healthcare for those with and without data

Accessibility/Usability of the technology, is there an age exclusion ?

Question 15

Ethical considerations with
wearables/phone data?

Answer 15

Privacy
Health inequalities
Biases

Question 16

Summarise bias as an ethical consideration with
wearables/phone data?

Answer 16

Does more expensive = better quality ?
Step count may only work for healthy gait
AI models is it trained or representative data

Question 17

Potential issues with wearables/phones?

Answer 17

Quality/Validation - designed for general population, will it pick up issues such as less pronounced steps, asymmetrical gait etc

Question 18

Data cleansing and formatting?

Answer 18

Outliers - is there anomalous data? excess steps = remove data or user

Variability - does it wildly vary weekly/monthly = remove user

Missing values - how many days missed, if any? = remove user

Question 19

Example of combining objective and subjective data?

Answer 19

Combine daily steps/change in step count (DV) with questionnaire variables eg age, gender (IV) to get context

Question 20

Equation for simple and multiple regression using step count example/ IV and DV?

Answer 20

Simple = y = B0 + B1x

Multiple = y = B0 + B1x1 + B2x2 + … + Bnxn

Simple = relationship between x and y

Multiple = relationship between y and all the xs
y = DV
x = IV
B0 = intercept

Question 21

Different classifications and regression types?

Answer 21

Class 0 = no/little % change
Class 1 = high % change

Continuous - regression
Dichotomous - logistics regression

Question 22

What can phones and wearables not tell us about a person?

Answer 22

Biomechanics

Question 23

Why are phones a good way to measure?

Answer 23

The ubiquity of ownership allows data collection at a large scale

Question 24

Potential bias of testing with phones?

Answer 24

Individuals without mobile phones, may cause an unrepresentative sample that may over-represent certain socioeconomic or age groups.

Question 25

Why is it important to check for, and possibly remove, inconsistent (highly variable) data in a dataset?

Answer 25

To ensure consistent and accurate results

Question 26

What is the primary difference between simple and multiple regression?

Answer 26

Simple regression has one independent variable, while multiple regression has multiple independent variables

Question 27

What would a significant odds ratio of 2.5 mean?

Answer 27

Those who are under that class have 2.5 higher odds compared to those who are not in the class