By Daphne Menheere, John Vlaming & Luc Hermans
In the 1950s Solomon Asch did a series of experiments to demonstrate the degree to which individual’s own opinions are influenced by those of a majority group. These experiments are now called Asch conformity experiments.
Inspired by the theory, it is interesting to see in what cases and how peer pressure influences peoples behavior. This project combines this idea with sporting activities. The lack of sporting and active movement is a global issue which makes it an interesting design topic. The research of this project looks into how peer pressure influences people to sport more.
A survey was conducted which resulted in 106 subjects who answered the questions. A data analysis with five assumptions created an interesting view on how non-sportsmen behave in contrast to individual and peer sportsmen.
The main hypothesis which will be tried to confirm is the following:
Do people sport more together then sporting alone?
After we formed our main hypothesis we had to decide in what way we wanted to gather our data. To be able to get a lot of data in a short period of time and directly in a digital format we decided to go for an online survey. We tested with SurveyMonkey and Google Drive. Because the Google Drive Form made in possible to save our data directly in .csv and .xlsx format and the interface was a lot more easy to use we decided to go for that option.
We tried to keep the survey as short and direct as possible so it would be easy to fill in, but at the same time gather more than enough data to draw our conclusions. We wrote down our main hypothesis and some sub-hypothesis and based our questions around these. We made a set-up of seven questions about the sport activities and besides that the age and gender of the participants.
Click here to go to the survey.
At first we gather the answers of 50 participants by walking through the Industrial Design spaces with a Macbook and iPad. Later on we decided to gather more data so we could draw more reliable conclusions. We did this by spreading the survey on Facebook and another walk around the spaces. We finally made our models with the data of 106 participants.
Click here for the datasheet: Data Sport Activities (106 Responses)
Most participants were at the age of 18 to 26. The gender ratio was quite evenly divided with 59% males and 41% females.
Within the data analysis, five different assumptions were questioned in order to confirm or reject our main hypothesis.
People who don’t sport have an equal priority level compared to the priority of people who sport.
To distinguish the non-sportsmen from the individual sportsmen and the peer sportsmen, participants answered the question ‘With whom do you sport?’ where multiple answers where possible: alone, in a team, with friends or I don’t do sports. This question made it possible to derive the different participants into categories: non-sportsmen, individual sportsmen and people who sport with peers (friends or in a team). Furthermore the participants were asked to prioritize their sport activities from a scale of one (low priority) till six (high priority).
By asking both questions, data was gathered that resembled the level of priority for the different categories: non-sportsmen, individual sportsmen and peer sportsmen.
See table X for gathered data
Table 1: Three different categories, which contain the priority levels.
This data is then stored into ILLMO in order to find the differences between the categories.
There are three different conditions (or stimuli) that are compared.
Stimuli 1: Non-Sportsmen
Stimuli 2: Individual Sportsmen
Stimuli 3: Peer Sportsmen
The first thing that needs to be done before the analysis can start is creating a model based on the gathered (imported) data. This is done by experimenting with the fitting of different models like Gaussian, Logistic, Laplace and Poisson (press the PDF box to choose a model).
When looking at figure 1 it is visible that different models had a different fitting as the LLC & AIC varied throughout the different iterations. Eventually the Laplace model had the lowest LLC & AIC and was therefore the most accurate model (figure 2). To create a better fit it is also possible to use BoxCox. However, this was irrelevant for this model, as this did not improve the fitting of the model.
Figure 1: Model history for fitting different models on data.
Figure 2: Eventual model (Laplace) used
With the eventual curve fitted model it is possible to begin the analysis. First, it is necessary to determine a Log-Likelihood Profile for the difference (LLP for all d(nref,n)) between stimuli 1 (non-sportsmen) against 2 (individual sportsmen) and 3 (peer sportsmen). This gives the following two graphs:
With these graphs it is possible to draw two conclusions. Within the first graph it is visible that the parameter value 0 is outside the two dotted lines. This resembles there is significant difference between the different stimuli. Within the second graph, stimulus 1 (non sportsmen) is taken as reference so the differences with stimuli 2 (individual sportsmen) and 3 (peer sportsmen) are visualized. As zero is outside the accuracy line of stimuli 2 and 3, it is possible to say that the priority of non-sportsmen is lower than the priority of sportsmen.
Within figure 3, stimulus 2 (individual sportsmen) is chosen as a reference and compared with stimuli 1 (non sportsmen) and 3 (peer sportsmen). As zero is outside the accuracy line of stimuli 1 and 3, it is possible to say that individual sportsmen have a lower priority than peer sportsmen.
Figure 3: Difference between individual sportsmen, against non-sportsmen and peer sportsmen.
The assumption can be rejected, as the priority of non-sportsmen is not equal to the priority of sportsmen.
Furthermore, this analysis showed the difference in priority for the different types of sportsmen: individual vs. peer sportsmen. Where peer sportsmen have a higher priority for sport activities than individual sportsmen.
People, who sport with peers, sport more than people who sport alone.
To confirm this assumption, the participants answered the question: ‘With whom do you sport?’ where multiple answers where possible: alone, in a team, with friends or I don’t do sports. This question made it possible to derive all sportspeople into categories: individual sportspeople and people who sport with peers (friends or in a team). Furthermore to know the total amount of time spending on sporting activities the participants answered the question: ‘How much time (minutes) do you spend on sport activities a week?’
So by asking both questions, data was gathered that resembles the total amount of time spending on sport activities for the different categories (individual vs. peer sportsmen).
This gathered data was imported within Illmo. However to start the data analysis, curve fitting was needed (see figure below).
The eventual curve fitting was reached by using the Logistic Model and BoxCox (overall power: -0,0612).
Figure 4: Model History Curve Fitting
Figure 5: Eventual Model
Figure 6: Probability Model
Figure 7: Thurstone Difference
By looking at the figure that resembles the Thurstone Difference, it is visible that the second dot and accuracy line does not cover the zero and there is a clear difference between the two dots.
Therefore it is possible to confirm the assumption where peers sportspeople sport more than individual sportspeople.
Sportsmen sport less then their perceived need of sport activities.
During the analysis of the data we were wondering how the perceived need of sporting activities related to the actual amount of sporting. The participants answered the questions: ‘How much time (minutes) do you spend on sport activities a week?’ and ‘In your opinion, what is the minimum amount of time (minutes) that you need to spend on sporting activities a week?’ in the survey. This way we could distinguish the perceived need of sporting activities and the actual amount of time spend on sporting activities.
Both data sets were entered as stimuli and as repeated measures.
Through trying out different probability density functions we found the best fitting model, the Student T function. Finally a BoxCox was added with a private variable for the overall power to make the model fit properly on the data histograms.
Figure 8: Model History Curve Fitting
Figure 9: Eventual Model
Figure 10: Probability Model
Figure 11: Thurstone Difference
Based on the Thurstone difference graph, we can say with a 95% certainty that the perceived need of sporting activities of sportsmen is lower than they actually sport. So the assumption can be rejected, as sportsmen sport more than their perceived need of sporting activities.
Non-sportsmen have a lower perceived need of sport activities than sportsmen.
To confirm or reject this assumption, the participants answered the question: ‘In your opinion, what is the minimum amount of time (minutes) that you need to spend on sporting activities a week?’ The answers resembled the perceived need of sport activities for the participants. To derive the non-sportsmen of the sportsmen the participants answered: ‘How much time (minutes) do you spend on sport activities a week?’
So eventually there were two stimuli based on the two categories (non-sportsmen and sportsmen) and their perceived need of sport activities. The eventual curve fitting was reached by using the Gumbell model, without the use of BoxCox.
Figure 12: Model History Curve Fitting
Figure 13: Eventual Model
Figure 14: Probability Model
Figure 15: Thurstone Difference
Even though there is a small difference between the two dots, the zero is within the accuracy range of the second dot. This is why it is possible to say that there is no significant difference between the two dots.
Therefore it is possible to reject our assumption and conclude that non-sportsmen do not have a lower perceived need of sport activities than sportsmen.
There is correlation between the perceived need of sporting time and priority to sport.
In order to confirm this assumption two questions from the survey are used to make the correlation. The following two questions were asked:
- How do you prioritize your sport activities?
- In your opinion, what is the minimum amount of time you think you need to sport?
In order to make a correlation between two variables in ILLMO a certain process need to be followed.
Make the amount of conditions the amount of data you want enter. In our case it is 106 (because of the 106 subjects).
When importing the data in ILLMO, the program asks if the data is repetitive. In this case you need to choose no and a dialog will open. The stimulus number is ‘line number’ and in order to see the correlation, we made the amount of minutes perceived, the column with data values and the priority level the predictor value. After that, hit ‘OK’.
Figure 16: Model and prediction pattern.
This is what it will look like. In order to see the correlation, click on ‘stats’ run a linear regression only test.
R = 0.181305
R confidence interval:
[ -0.010639, 0.360363]
Figure 17: Linear Regression Only plot.
When running the test it is possible to see that there is almost no correlation between the priority and amount of time they think they need to sport. Furthermore, zero is within the confidence interval that resembles that it is also possible there is no correlation at all. So the assumption can be rejected, as there is almost no correlation between the perceived need of sporting time and priority to sport.
Based on the data analysis our main hypothesis can be confirmed: people who sport together (with friends or in a team) sport more than people who sport alone.
Different design opportunities rose during the data analysis.
We did not expect a similar perceived need of sport activities for sportsmen and non-sportsmen. The interesting thing about this is that many interventions focus on raising the awareness of the importance of sporting activities among people, while our research already resembles a certain awareness for people who sport and don’t sport. So for a new design case/intervention the focus would not be on raising awareness of the importance of sporting but increasing their priority to do sports.
Furthermore, by investigating the correlation between the perceived need of sport activities and priority to sport we could conclude there is no correlation. For us this resembled that people have awareness in how much they need to sport but don’t have it high on their list of priorities.
Another interesting design opportunity we could derive from the research is that peer sportsmen have a higher priority level for sport activities and sport more than non-sportsmen and individual sportsmen. This makes it interesting to motivate non-sportsmen to start sporting with friends or in a team.
During the data analysis we found out that our approach had some discussable aspects. The first discussion point is that we had no independent variables in our data, except for the age and gender. All the data sets were the interpretation or thoughts of people, and we had no measured data.
The question in our survey were nog fully in line with each other. We used different categories in different questions, like the separation between fixed and non-fixed in one question, but the separation between alone, with friends, or in a team in another question. Therefore we could not compare all data without making some minor assumptions.
We asked quite a lot of people for our survey, but the participants do not reflect the whole population of the Netherlands. Almost all the participants are between the ages of 18 to 26, and most participants are Industrial Design students.