Quantified self Individual Report for DSI AT2
13484528
Vu Viet Truong (Vincent)
24/10/2019
0.1 Introduction
Quantified self-tracking is to monitor and track human-related data such as biological, physical, behavioral or environmental information (Swan, 2009). Simply, it is to use self-tracking technologies to monitor and measure many aspects of human’s life, such as steps, calories, sleeping hours, emotions, stress levels and so on. Two principles often come together with quantified self-tracking are ‘self-knowledge through numbers’ (Ruckenstein, 2015; Schüll, 2016;Swan, 2012) and ‘self-optimisation’(Choe et al., 2014:1147;). There could be an endless list of things to measure but QS concentrate on finding a meaningful insights of QS data and leverage that knowledge to change and improve human life.
Health contributes a major reason for quantified self tracking but some execute QS to share their common interest, achieve health goal, explore an idea or solely be curious (Swan, 2013). Self tracking may start at the individual level but it is quickly extending to group level where trackers share and work collaboratively with their QS data. Several community groups has been created to share their best practices in developing sustainable self tracking activities, including Quantified Self Community comprised 70 meetup groups with 5,000 participants (http://quantifiedself.com/) or Habit Design (www.habitdesign.org). With that in mind, we as a group of three came to design our quantified self experiment project to explore general idea, knowledge, tools, techniques and experiences.
To explore and understand QS extensively, a variety of areas were tracked and analyzed in this experiment. Health is important but not exclusive focus, where our objective ranges from physical activities to psychological and social enhancement. However due to limitation of two-month project, it is designed with an ease for students and broad understanding of QS but towards to main objective, which is to explore the changing emotions of by effects of physical health, measured by steps and location, or social forces such as call log and weather as environment influences. We also apply the same standard format (Swan, 2013) as many QS projects, which are presented in a simplified version of the scientific method, answering three questions:
What did you do? We investigate the effects of range of variables on changing human emotions.
How did you do? How limitation of tracking tools or bad practices of member could affect data quality issues in quantified self tracking?
What did you learn? Through investigating the relationships in data, it could provide insights into one’s own life.
A few discussions about data privacy and ethics is included as an extended part. It is necessary to clarify privacy standards for how personal data is used in all QS experiments, especially in experiments with sensitive personal information of group members.
0.2 Description of process or methods
With a design for understanding multifacets of QS tracking, we approach range of variables, from physical activities, psychological to social and environmental aspects. Inspired by a project named Sen.se, an experiment to find linkage between social interaction and mood (Swan, 2013), we take emotions as our outcome value and try to find the correlation with multiple variables as steps or weather. There are few articles provides evidence about a positive relation between short walks and mood (Ekkekakis, Hall, et al., 2000), or weather and mood (Denissen et al., 2008) which is a good reference.
Summary of our datasets, measures and method
| Type | Dataset | Measures | Method |
|---|---|---|---|
| Group | Steps Count | Daily number of steps | Google Fit |
| Group | GPS Location | Location, distance | Google Location |
| Group | Selfie photos & Emotions | Emotions level | Camera & Advanced AI software |
| Indivdual | call log | Time & Call duration | Mobile phone & Phone app |
| Open | Weather | Daily temperature | Downloaded from website |
At first, we discussed a couple of tools to collect steps count and gps locations but we finally end up with Google Fit and Google Location due to a few reasons. With Google Location, it was available on every member’s phones so we only need to turn on location settings on our phone. For steps, we need to download and install Google Fit. Data frequency is not our concern because both apps collect and store data seamlessly into individual Google cloud account so all we have to do is to carry our phones wherever we go. At the end of the experiment, each member downloaded data from their Google account and sent to other members.
We leveraged power of one advanced AI system to measure emotions and mood through our daily selfie photos. So everyday, selfie photos were taken and stored by each member. At the end, we collected photos and shared to group via Google drive. There were few options for advanced AI software as Feely app (https://play.google.com/store/apps/details?id=com.vladimir.feely&hl=en), Microsoft API (https://azure.microsoft.com/en-au/services/cognitive-services/face/) and Sightcorp (https://api-face.sightcorp.com/api/detect/). After trying a couple of options, we went for Sightcorp solution. It provides free trial API services for two weeks, which is sufficient for us to do our analysis. However, a limitation of Sightcorp or Microsoft is that all photos need to be hosted on server to get the API run properly. So we uploaded photos onto https://imgbb.com/ server and wrote few lines of python code to run API through all of images then wrote it in csv format file. Sightcorp can identify 7 basic emotions based on the position and movement of facial muscles. Each analysis produces a confidence score for different types of emotions: anger, disgust, fear, happiness, surprise, sadness, and neutral.
Since this is the only dataset to collect manually, the frequency of taking photos is not as expected. At first, we agreed that we will share our selfie photos everyday on one Whatsapp group chat. However, few were not comfortable with this share. We chose to store photos individually and share at the end. As few students may forget to take photos on a few days so we end up with a considerable amount of missing data.
We took call log and weather as individual and open datasets, respectively. Callog is collected and stored automatically on my phone. To extract call logs from my phone, I used the app “SMS, Call - XML, PDF, CSV(Super Backup & Restore)” (https://play.google.com/store/apps/details?id=com.greenchills.superbackup&hl=en_AU). This data is quite sensitive so I removed the phone number and change few caller names to anonymous. Weather data was downloaded from the “National Centers for Environmental Information” (https://www.ncdc.noaa.gov/cdo-web/results).
As principle of data ethics, we did anonymise appropriately before doing analysis. So person 0, 1, 2 will be used to identify three members in this report.
1 Analysis
Data summary
According to date of three datasets, we chose to go from range from 2019-Aug-05 to 2019-Sep-14 even though location data of few members were given with few years of duraton. Below is summary of datasets:
## location emotions steps
## 1 12745 43 40
## 2 1033 98 41
## 3 20825 101 41## <BarContainer object of 3 artists>## <BarContainer object of 3 artists>## <BarContainer object of 3 artists>
Amount of data for emotions and locations are different between members while steps are quite sufficient among the three. It is understandable that the missing data on emotions because members could forget to take selfie photos for a few days. However, location is collected automatically by Google app, we still have missing data from Person 0 and 1.
Check data frequency
We investigate the frequency of these two datasets, location and emotions as below:
At this angle, it gives details of missing locations datasets of person 0 and 1. Person 1 did not seem to take locations during the first two weeks while the person 2’s missing data is random. He may forget to bring phone on few days.
Same as location, frequency of emotions data of person 0 and 1 are quite random while person 2 is nearly perfect, which has only one missing around 14 Aug. Another point is that previous chart shows the same total amount of photos between person 1 and 2 but frequency of person 1 was given worse.
Steps analysis
The average step of Person 0 and Person 2 are about 7000 steps / day. It is nearly match with (Tudor-Locke et al., 2011)’s recommendation, which is 8,000 daily steps and 7,100 steps/day. There are two outliers on the graph, which could be a day trip of person 0 and person 1 during that period. On average, person 0 and 2 are likely to be more active than person 1.
Plotting average steps per weekday, we can see that Monday, Tuesday, Saturday and Sunday are the busiest days of three because we are having class on that day. Sunday is also the same because it is a weekend. It is remarkable that Person 0 made a considerable amount of steps every Saturday.
Locations analysis
## Figure(layout=FigureLayout(height='420px'))## Loading required package: ggplot2## Google's Terms of Service: https://cloud.google.com/maps-platform/terms/.## Please cite ggmap if you use it! See citation("ggmap") for details.## Source : https://maps.googleapis.com/maps/api/staticmap?center=Sydney&zoom=11&size=640x640&scale=2&maptype=hybrid&language=en-EN&key=xxx-gXQYSA## Source : https://maps.googleapis.com/maps/api/geocode/json?address=Sydney&key=xxx-gXQYSA## Warning: Removed 816 rows containing missing values (geom_point).
Looking at the heatmap, highlight locations are Sydney CBD, Chatswood, Hurstville, Strathfield. Sydney CBD is certainly a focus point of the three because they were studying at the same university in Sydney CBD. The other highlighted locations could be the member’s house.
## Source : https://maps.googleapis.com/maps/api/staticmap?center=Sydney&zoom=11&size=640x640&scale=2&maptype=satellite&language=en-EN&key=xxx-gXQYSA## Source : https://maps.googleapis.com/maps/api/geocode/json?address=Sydney&key=xxx-gXQYSA
The path map presents members’ moving from location to location. Few straight lines from Parramatta to Sydney CBD, Strathfield to CBD, Chatswood to CBD could give more evidence to their home location.
Now, we take a detailed analysis to see how frequent they were moving on a weekday basis.
On the above graph, Person 0 owns the highest average moving, 100 km and 50 km, on Tuesday and Sunday, respectively. Apparently, it can be predictable that person 1 has the lowest average of moving everyday, except Saturday and Sunday. It is also abnormal that he had no moving on Wednesday or we may be given the missing data. Meanwhile, Person 2 made about the same average moving everyday, except quite low on Sunday.
Emotions analysis
Given quite considerable missings of emotions data, we do not make comparison and contrast between group members but a whole of data will be combined and analysed.
Sightcorp can identify 7 basic emotions based on the position and movement of facial muscles. Different types of emotion are given by a confidence score from 0 to 100, in which the greater score, the likely of emotion would be.
## emotions-anger emotions-disgust ... emotions-sadness emotions-surprise
## 14 3.0 10.0 ... 39.0 4.0
## 15 2.0 9.0 ... 24.0 11.0
## 16 4.0 5.0 ... 25.0 15.0
## 17 3.0 5.0 ... 52.0 2.0
## 18 1.0 4.0 ... 36.0 10.0
##
## [5 rows x 6 columns]Given the confidence score logic of Sightcorp, the highest score on one day likely to presents the emotion state of that person on that day. So we strip off low confidence score and remain the highest one to serve as the status of member as whether they were angry, fearful, happy or surprised.
## <BarContainer object of 6 artists>## ([<matplotlib.axis.XTick object at 0x1218f4ef0>, <matplotlib.axis.XTick object at 0x1218f4828>, <matplotlib.axis.XTick object at 0x1218f4128>, <matplotlib.axis.XTick object at 0x1203efc50>, <matplotlib.axis.XTick object at 0x120402358>, <matplotlib.axis.XTick object at 0x120402eb8>], <a list of 6 Text xticklabel objects>)
Happiness is quite low while sadness is at the peak in terms of appearances during the period. However, the average level of happiness is highest, over 40 while over 20 is an average presence of sadness. It could be explained by that the pressure of high living cost could cause the sadness while they may be exhilarated with a few days of Sydney discoveries.
Calllog analysis
## name call_time call_log duration
## 156 Lawrence LAM 2 2019-09-13 20:50:00 Dialed 3.90
## 157 Lawrence LAM 2 2019-09-13 20:46:00 Dialed 0.03
## 158 Ganesh UTS 2019-09-13 20:46:00 Dialed 0.00
## 159 Lawrence LAM 2 2019-09-13 20:41:00 Missed 0.00
## 160 Lawrence LAM 2 2019-09-13 20:39:00 Missed 0.00
## 161 Ganesh UTS 2019-09-13 20:26:00 Received 15.12
## 162 Ganesh UTS 2019-09-13 20:25:00 Received 0.00
## 163 Abhisek UTS 2019-09-13 20:17:00 Dialed 7.83
## 164 Lawrence LAM 2 2019-09-13 19:55:00 Received 13.93
## 165 Ganesh UTS 2019-09-13 15:59:00 Received 1.92
Violin plot shows that duration of call I received is higher than I made, in which PM calls were higher than AM calls. Additionally, there were few missed and rejected calls during that period.
Given average duration of call logs during the day, 9 PM weekdays makes an exceptional high with 20 minutes. Another surprise is that, 12 PM is quite late at night but the call log duration is quite high on both weekends and weekdays. Generally, calls were made more on weekdays than weekends.
Zooming in the 9 PM weekdays, the boxplot presents the two outliers, 43 and 25 minutes duration of calls, that could explain why it is remarkably high. It could be the day we have a long group call to do my assignment.
## (-0.5, 299.5, 199.5, -0.5)
After applying wordcloud on our data, old friends, family and UTS friends were the subjects I called primarily.
Weather analysis
## station name date tavg tmax tmin
## 0 ASN00066037 SYDNEY AIRPORT AMO, AS 2019-08-05 14 21 9
## 1 ASN00066037 SYDNEY AIRPORT AMO, AS 2019-08-06 13 21 7
## 2 ASN00066037 SYDNEY AIRPORT AMO, AS 2019-08-07 15 23 7
## 3 ASN00066037 SYDNEY AIRPORT AMO, AS 2019-08-08 16 22 8
## 4 ASN00066037 SYDNEY AIRPORT AMO, AS 2019-08-09 13 18 13
During this period, the weather is quite cold to us because we came from tropical countries to UTS to study. The average temperature from less than 10 to around 20, which is quite low, even without taking into account the high wind on few days, which could make it even harsher. Additionally, diverse range of temperature during the day and among the period could also be contribute to savage environment conditions during winter.
Correlation between variables and emotions
Person 2 appears to be the one who made selfie photos regularly and his data is perfect in terms of quality and frequency. So his dataset is chosen to take into the correlation analysis between emotions and other variables.
First thing first is to merge by date all datasets of emotions, steps, distance retrieved from locations, temperature.
## date emotion emotion_value steps distance tavg
## 0 19-08-05 emotions-fear 15.750000 NaN 25.02 14
## 1 19-08-06 emotions-sadness 19.090909 5121.0 32.46 13
## 2 19-08-07 emotions-sadness 14.000000 575.0 4.40 15
## 3 19-08-08 emotions-sadness 19.692308 7152.0 48.65 16
## 4 19-08-09 emotions-sadness 16.400000 2689.0 27.47 13
## 5 19-08-10 emotions-sadness 30.500000 8937.0 33.52 12
## 6 19-08-11 emotions-sadness 41.125000 7709.0 18.95 12
## 7 19-08-12 emotions-anger 19.500000 6438.0 6.64 12
## 8 19-08-13 emotions-sadness 23.642857 12638.0 38.25 12
## 9 19-08-14 emotions-sadness 21.400000 10407.0 13.21 13On the merged datasets, emotion again is extracted to retain only the highest confidence score on each day while steps and distance are aggregated on a daily basis.
Amount of sadness emotion dominates the emotions board so it could play as an indicator representing the member’s emotions. That way could convert the emotion into simpler pattern when putting it into correlation analysis, where the high score of sadness presents sadness and the lower score presents less sadness.
## date emotions_sadness steps distance tavg
## 0 19-08-05 13.000000 NaN 25.02 14
## 1 19-08-06 19.090909 5121.0 32.46 13
## 2 19-08-07 14.000000 575.0 4.40 15
## 3 19-08-08 19.692308 7152.0 48.65 16
## 4 19-08-09 16.400000 2689.0 27.47 13
Plotting all variables to find missing values, we found steps dataset missing a couple of data entry. We fill these missings with mean value.
## <seaborn.axisgrid.PairGrid object at 0x121e17978>
Plotting all variables on a scatter plot, at the first row, we found that there is no correlation between sadness emotion with steps and distance. There is low negative linear relationship between emotion with average temperature.
However, plots give a certain positive correlations between steps and distance. Put differently, it could disclose the further distance students made, the more steps he must take to reach his desired destination. It is quite obvious because this student has no car and mostly travel by public transport.
## [Text(0, 0, '19-08-05'), Text(0, 0, '19-08-06'), Text(0, 0, '19-08-07'), Text(0, 0, '19-08-08'), Text(0, 0, '19-08-09'), Text(0, 0, '19-08-10'), Text(0, 0, '19-08-11'), Text(0, 0, '19-08-12'), Text(0, 0, '19-08-13')]
Time Series plot does not assist much to find the same pattern among variables, except it exists some signs of the same trending between steps and distance.
2 Findings and conclusion
Steps dataset is the most qualified dataset in terms of frequency and accuracy while it is not the case for location and emotions. One person was doing his excellent job by giving all pretty qualified datasets.
Two persons’ daily steps count are given as about 7000, which approximately match with (Healthline, 2019)’s recommendation. The steps reveals little higher count on a few studying days and weekends. In relation with steps, location shows that three persons moved around Sydney, specifically UTS university and home.
Spending more time for call during PM than AM, specifically 9 PM and 12 PM is what we found in individual call logs. Obviously, calls were made mainly to friends, family members living in Sydney. Meanwhile, temperature is normally low at that period, given by weather dataset.
Scatter plotting all variables shows no strong correlation between emotions and other variables, except positive correlation between steps and distance. As mentioned earlier, there exists a number of experiment projects which could identify few signs of relationship. Here, we think of a few possible reasons for this such as: - Quality of data collection is major cause: missing data, missing appropriate of frequency. - One or two months data or one or two person may not provide a sufficient amount of data. - Selecting granularity of day could be inappropriate such as emotions could change during the day
3 Discussion
Data regulation
Increasing availability of QS tracking devices, wearable devices and phone apps could bring benefits for both individual and business in QS activities. However, it emerges threats of privacy issues which should be addressed through technology and regulatory mechanisms (Brown, Brown and Korff, 2010). Given the strategic role of regulatory mechanism in dealing with such issues, a certain number of countries have enacted privacy laws which seek to protect individual information such as: CCPA (California Consumer Privacy Act), GDPR (EU’s General Data Protection Regulation) and so on. In this regard, privacy issues is our primary concern in this experiment. GDPR regulation is taken into account to assess all of our activities including data collection, data analysis and third party tools application to evaluate its regulatory compliance.
Feature 2 of GDPR proposes that Data Protection Impact Assessments (DPIAs) should be conducted in a plan to track individuals’ location or behavior while Feature 3 suggests the integration of privacy into design, processes and procedure (Paruch et al., 2019). We have not conducted thorough DPIAs at our beginning steps but we were mindful our data privacy and our risk mitigation for better protect member’s information. One example is that when one of our members felt uncomfortable in giving out his call logs data, we removed this from our group dataset. We contemplate third party tools and select one which comply with data protection laws. Tools as (Sightcorp, 2019) and (OAIC, 2019) state their GDPR compliance on their website.
Conforming with Feature 5, Data Subject Access Requests (DSAR), we only collect and analyse data within what we agreed on. Even two members gave out their few years of duration of locations data but we did remove all redundant records from our data collection.
Regardless of our mindful data privacy consideration, we know that it may exist a few inevitable privacy issues. The two tools we were using to host images, imgbb website (https://imgbb.com/privacy) and extract our individual call logs, Super Backup app (https://play.google.com/store/apps/details?id=com.greenchills.superbackup&hl=en_AU) do not state explicitly their privacy regulations. We accepted this risk as we have no better choice but we mitigate our risks by removing all our data after completing necessary tasks.
Data ethics
Data ethics is defined as an action-oriented analytical framework or proactive agenda which could aim to balance powers embedded in the Big Data Society and drive towards the human-centric distribution of power (Gry, 2019). Simply put, it is about responsible and sustainable use of data or make it right for people and society. It requires a holistic approach includes principles and guidelines in data processing activities (GOV.UK, 2019). Five components of principles of data ethics proposed by (Pernille, 2018), will be piped through our experiment to evaluate its ethical standards.
The human being at the centre: Our experiment is designed to put primary benefits to our group members, which is clearly stated in our goal, which is to explore and understand member’s insights through QS experiment.
Individual data control: All data were recorded individually and committed orally to use within boundary of assignment.
Transparency: All members understand the risk as well as societal consequences. It is only two-months project and all members are students with the main activities involving around studying so it should not cause high risk and ethical consequences. However, de-identification and choice of less-sensitive data could mitigate the risk at minimum.
Accountability: Sustainable personal data processing such as data anonymisation is taken into account to reduce the risks for the individual and mitigate social and ethical implications.
Equality: We are students studying at the same university so we believed that we have the same societal power. It could reduce risk of being discrimination or stigmatisation in our judgement.
4 Reflection
Through the journey of this experiment, we appreciate the support of the members, instructors and friends to help us accomplish this project. The most difficult part to me is the beginning of the project, when we almost have no idea what is QS, how to set up a QS project running, which data and how to track and analysis, what is the goal. Through many workshops, meetings, studying and reading, we push the progress forwards step by step. What we achieved so far is the better understanding QS, how to set up goals, format for QS project and how to run it. However, there are few backwards and failures which need to address to turn it into lesson learnt for future experiment.
We are quite unsure why we cannot find the correlation between emotions and few other variables while project Sen.se could find apparent linkage between mood and coffee consumption, social interaction (Swan, 2013). (Roberts, 2004) found influences to mood by morning TV evening faces and morning light. Compared with Roberts, we found two major differences between our experiment and his. First, he made his own scale of mood while we used Sightcorp advanced AI tools. Having limited time to test and verify AI algorithm on our selfie photos, the quality of this tool remains a question to us. Second, the granularity of Rober’s mood records is quite dense, an interval of few hours during the day while our scale is once a day. It may be the root cause because person could change mood during a day so the mood and steps should be recorded and analysed at a relative point.
Regarding to data ethics and privacy, the first thing we could do better is to apply the GDPR’s Data Protection Impact Assessments (DPIAs) and Privacy by Design (PbD) at the beginning of our project (Paruch, 2019). We could follow template given by the UK’s Information Commissioner’s Office (ICO) as a guideline through the process. PbD should be the best practices for every QS project to fully integrated data privacy and security into the design processes, procedures, protocols, and policies of a data processing. Additionally, principles of ethics framework by (GOV.UK., 2019) should be considered to perform to get ethics right and achieve the highest standards for transparency and accountability when processing data privacy. Lastly, it is critical to set our mind to nominate the third party tools who states GDPR compliance officially on their website on every circumstance.
5 References
Brown, I., Brown, L., and Korff, D. Using nhs patient data for research without consent. Law, Innovation and Technology 2, 2 (2010-12-01T00:00:00), 219–258.
Denissen, J., Butalid, L., Penke, L. and van Aken, M. (2008). The effects of weather on daily mood: A multilevel approach. Emotion, 8(5), pp.662-667.
Ekkekakis, P. & Petruzzello, S. J. (2000c). Analysis of the affect measurement conundrum in exercise psychology: III. A conceptual and methodological critique of the Subjective Exercise Experiences Scale. Psychology of Sport & Exercise, 2, 205-232.
Gary Wolf and Ernesto Ramirez (2014). Quantified Self Public Health Symposium April 2014
GOV.UK. (2019). Data Ethics Framework. [online] Available at: https://www.gov.uk/government/publications/data-ethics-framework/data-ethics-framework [Accessed 24 Oct. 2019].
Gry Hasselbalch (2019), Making sense of data ethics. The powers behind the data ethics debate in European policymaking, Volume 8, Issue 2 PUBLISHED ON: 13 Jun 2019 DOI: 10.14763/2019.2.1401, Department of Information Studies, University of Copenhagen, Denmark
Norris J2012Self-Tracking May Become Key Element of Personalized Medicine www.ucsf.edu/news/2012/10/12913/self-tracking-may-become-key-element-personalized-medicineMarch202013.3. Norris J. 2012. Self-Tracking May Become Key Element of Personalized Medicine. Available online at www.ucsf.edu/news/2012/10/12913/self-tracking-may-become-key-element-personalized-medicine (Last accessed on March 20, 2013). Google Scholar
OAIC. (2019). Guide to data analytics and the Australian Privacy Principles. [online] Available at: https://oaic.gov.au/privacy/guidance-and-advice/guide-to-data-analytics-and-the-australian-privacy-principles/ [Accessed 24 Oct. 2019].
Paruch, Z., Team, T., Sebastian, F. and Dearie, K. (2019). GDPR Compliance | Everything You Need to Know | Termly. [online] Termly. Available at: https://termly.io/resources/articles/general-data-protection-regulation-gdpr-compliance-guide/#gdpr-compliance-requirements [Accessed 24 Oct. 2019].
Pernille Tranberg, Gry Hasselbalch, Birgitte Kofod Olsen & Catrine Søndergaard Byrne (2018) DATAETHICS – Principles and Guidelines for Companies, Authorities & Organisations, Isbn print: 9788771920475 Isbn pdf: 9788771920482 Isbn epub: 9788771920499, Available at: https://dataethics.eu/wp-content/uploads/Dataethics-uk.pdf
Roberts, S. (2004). Self-experimentation as a source of new ideas: Ten examples about sleep, mood, health, and weight. Behavioral and Brain Sciences, 27(02).
Sightcorp. (2019). Sightcorp Software Privacy Features | Sightcorp. [online] Available at: https://sightcorp.com/privacy-features/ [Accessed 24 Oct. 2019].
Swan, M. (2013). The Quantified Self: Fundamental Disruption in Big Data Science and Biological Discovery. Big Data, 1(2), pp.85-99.
Swan, M. (2009). Emerging Patient-Driven Health Care Models: An Examination of Health Social Networks, Consumer Personalized Medicine and Quantified Self-Tracking. International Journal of Environmental Research and Public Health, 6(2), pp.492-525
Tudor-Locke, C., Craig, C., Aoyagi, Y., Bell, R., Croteau, K., De Bourdeaudhuij, I., Ewald, B., Gardner, A., Hatano, Y., Lutes, L., Matsudo, S., Ramirez-Marrero, F., Rogers, L., Rowe, D., Schmidt, M., Tully, M. and Blair, S. (2011). How many steps/day are enough? For older adults and special populations. International Journal of Behavioral Nutrition and Physical Activity, 8(1), p.80.
6 Appendices
df_slice['location'].head(10)## date distance latitude ... day day_name day_time
## 0 2019-08-17 14:47:21 0.00 -33.910375 ... 5 Saturday PM
## 1 2019-08-17 14:49:22 1.29 -33.911948 ... 5 Saturday PM
## 2 2019-08-17 14:51:24 1.18 -33.909619 ... 5 Saturday PM
## 3 2019-08-17 14:53:26 0.77 -33.912471 ... 5 Saturday PM
## 4 2019-08-17 14:55:28 1.03 -33.917313 ... 5 Saturday PM
## 5 2019-08-17 14:57:21 1.20 -33.919962 ... 5 Saturday PM
## 6 2019-08-17 14:59:30 0.85 -33.922793 ... 5 Saturday PM
## 7 2019-08-17 15:01:38 1.10 -33.925608 ... 5 Saturday PM
## 8 2019-08-17 15:03:39 1.75 -33.917929 ... 5 Saturday PM
## 9 2019-08-17 15:05:41 0.47 -33.917789 ... 5 Saturday PM
##
## [10 rows x 11 columns]df_slice['emotions'].head(10)## age clothingcolors-1 clothingcolors-2 ... day day_name day_time
## 14 38.0 #111111 #111122 ... 1 Tuesday PM
## 15 38.0 #110000 #111111 ... 3 Thursday AM
## 16 42.0 #221111 #331111 ... 4 Friday PM
## 17 36.0 #aaaaaa #bbbbbb ... 5 Saturday AM
## 18 31.0 #000000 #000011 ... 6 Sunday AM
## 19 37.0 #888855 #886666 ... 6 Sunday PM
## 20 38.0 #cccc99 #ddddaa ... 6 Sunday PM
## 21 40.0 #bbbb88 #cccc99 ... 6 Sunday PM
## 22 34.0 #bbbb77 #bbcc88 ... 6 Sunday PM
## 23 NaN NaN NaN ... 6 Sunday PM
##
## [10 rows x 37 columns]df_slice['steps'].head(10)## Average speed (m/s) Average weight (kg) ... day_name day_time
## 1 NaN NaN ... Monday AM
## 2 1.259258 65.0 ... Tuesday AM
## 3 0.780275 NaN ... Wednesday AM
## 4 1.041264 NaN ... Thursday AM
## 5 0.986473 NaN ... Friday AM
## 6 1.272918 NaN ... Saturday AM
## 7 0.864163 NaN ... Sunday AM
## 8 1.272006 NaN ... Monday AM
## 9 1.394393 NaN ... Tuesday AM
## 10 0.847609 NaN ... Wednesday AM
##
## [10 rows x 27 columns]df_calllog.head(10)## name call_time call_log ... day_name weekday weekend
## 156 Lawrence LAM 2 2019-09-13 20:50:00 Dialed ... Friday 4 False
## 157 Lawrence LAM 2 2019-09-13 20:46:00 Dialed ... Friday 4 False
## 158 Ganesh UTS 2019-09-13 20:46:00 Dialed ... Friday 4 False
## 159 Lawrence LAM 2 2019-09-13 20:41:00 Missed ... Friday 4 False
## 160 Lawrence LAM 2 2019-09-13 20:39:00 Missed ... Friday 4 False
## 161 Ganesh UTS 2019-09-13 20:26:00 Received ... Friday 4 False
## 162 Ganesh UTS 2019-09-13 20:25:00 Received ... Friday 4 False
## 163 Abhisek UTS 2019-09-13 20:17:00 Dialed ... Friday 4 False
## 164 Lawrence LAM 2 2019-09-13 19:55:00 Received ... Friday 4 False
## 165 Ganesh UTS 2019-09-13 15:59:00 Received ... Friday 4 False
##
## [10 rows x 9 columns]