Streetscapes and Walkability
October 28th, 2010, by Jeremy Krygsman
This section of the report will present the data that was collected from the conducted primary research. Both quantitative and qualitative data will be represented. Again, it is the purpose of this report that greenery will have significant impact on a street scape. It is for this reason that we have created a hypothesis that will be tested through our statistical analysis.
Null hypothesis: Greenery will not have a significant impact on participants.
Hypothesis: Greenery will have a significant impact on participants.
To begin, the following graph depicts ratios of male to female, as well which video had been chosen as the preferred alternative. To reiterate, video titled level 1 represents the lowest quality of greenery. This video depicts a cemented streetscape without any vegetation, and thus it is the control and basis of comparison for all other visualizations. The video titled level 4 represents the highest quality of greenery as it includes shrubs, flower boxes, and trees. A total of 38 surveys were conducted, providing both quantitative and qualitative responses. Of the total, the ratio of male to female surveyees was 21: 17. Overall, this is relatively a balanced distribution amongst both genders. Table 2 illustrates how often each visualization was selected as the preferred option. Evidently the visualization with the highest inclusion of vegetation (level 4) had been chosen the most often.
|Quality of Streetscape||Frequency as Preferred Option|
|Level 4 (highest inclusion of vegetation)||27|
|Level 1 (Lowest inclusion of vegetation)||2|
|*Other, indicated two top choices||1 (chose both Level 3 and 4)|
|*N/A, did not answer question||2|
Figure 1 illustrates the difference in frequency among all the visualizations being chosen as the preferred option, based on their level of greenery. Clearly, level 4 exceeds the frequency count of all other levels, as it was selected 27 times as the preferred option. Thus level 4 had been chosen 27 out of 38 times, giving this visualization a 71% of being chosen as the preferred alternative. It is important to note, however, that one surveyee chose two options as their preferred alternative, being both the visualizations of highest greenery (level 3 and 4). Furthermore, two other participants had not answered the question, as is included the table 1 above. Removing these three responses allows level 4 to be chosen 27 out of 35 times, creating a percentage of 77 % as the preferred option. Also of note, levels 1, 2 and 3 are evenly distributed in frequency, as level 1 and level 3 were both chosen as the preferred option 3 times (each). This depicts a high contrast between levels of greenery further showing that the highest level of greenery was the preferred option.
Figure 2 represents how often each visualization level had been ranked, by using a Likert Response scale. This type of scale was used in order to facilitate the answers of participants and to aid data analysis by using interval level measurement that allows easy comparison. It is important to state that the survey question provided a ranking system of 1 to 5, 1 being „very undesirable,‟ 3 being „neutral,‟ and 5 being „very desirable.‟ Figure 2 depicts how often each visualization received a ranking between 1 and 5. The control video (level 1) received a ranking of very undesirable approximately 2 times, a ranking of neutral 20 times, and a ranking of very desirable 0 times. In contrast, level 4 did not receive a ranking below neutral, again proving it to be the most desired option. To reiterate, level 4 is the alternative which was ranked very desirable approximately 17 times.
Figures 3 and 4 are also based on a Likert scale, which allows the evaluation of safety for each visualization during both day and night. Comparing the two graphs, visualizations including vegetation received higher rankings in the day for safety, than at night. The significant finding is that visualization 4, though it had received rankings of neutral and above, had received mixed reviews of safety at night time, expressed by the ranking of undesirable (seen in Figure 4).
The figures below (5 and 6) represent how visualizations compared in safety to the control video (being level 1, with bare concrete). According to the median and the mode, levels 3 and 4 were the preferred option in comparison to level 1. Levels 3 and 4 have the same median and mode in regards to safety during the day, however at night these statistics do not remain the same. At night, level 3 received the same ranking of safety as that to level 1 (for the mode and median). As for levels 2 and 3 at night, these two levels are perceived to have the same level of safety as level 1 (in regards to the mode and the median). Though these three statistics are provided, the mode would be the most important indicator for comparison of safety. As is evident from the graphs, level 4 received the greatest ranking for safety in both day and night.
Figure 7: Safety during the day
Figure 7 illustrates the mean difference between each visualization compared to the control level. Evidently, level 4 is selected often as the safest option of all four videos. This perception, however, does not hold true for 3 participants. Interestingly, level 2 received the weakest perception in safety as it appears 6 times below the mean. On the other hand, level 2 appears at 2.00 a total of four times, thus for some participants this level of vegetation was seen as safe. Level 3 is seen as being unsafe, as it too is depicted being below the mean a total of 4 times, however it appears above the mean a total of 23 times. Of these 23 times, level 3 is depicted above 1.00 only 4 times.
4.1 Statistical Tests
Data was tested in order to define whether there was a normal distribution, however, using SPSS, results indicate that the data was non-normal. Reasoning for the non-normality of the data relates to the sample size of 38 for the conducted surveys.
This table below represents the results of the Shapiro-Wilk test. Since the p-value (indicated by Sig.) is lower than 0.5, then the data is non-normal. This is assumed due to the small scale that was used throughout the survey (1 to 5 scale) as well as the small sample size.
4.1.2 Friedman’s Anova
The Friedman‟s Anova test was used in order to look at assumptions of normality (and because the data is non-normal). By using box plots, the standard deviations were examined for each visualization. If the standard deviations do not overlap, then the data becomes more significant. In fact, as the standard deviation is tighter then the significant difference increases. The Friedman‟s Anova test indicates whether there is a great significant difference amongst all the four levels, however does not identify between which levels the difference occurs. Evidently, there is a significant difference amongst the four levels, as presented by the box plots below:
Figure 8: Level of Greenery Friedman’s Anova Test
The question regarding safety used the Likert scale, in which participants could rank on a scale of 1-5. The standard deviations of these separate box plots do not overlap, causing the data to be ever more significant. Since the standard deviations of all visualizations do not overlap, this means that there is a difference between the levels of greenery. The limitation of this knowledge is that the Friedman‟s Anova test does not indicate where this difference occurs. In order to determine where (or at which level of greenery) the greatest difference occurs, the Wilcoxon test will be used. By interpreting the graph, visualization 4 (with the highest level of greenery) had the greatest perception of safety during the day.
Furthermore, the box plots illustrate that there are outliers within the data set. The skewness of the data is presented by the median line not being at equidistant from the inter quartile ranges.
Still using Friedman‟s Anova, comparison of box plots were computed for the perceived safety of both night and day. The box plots for safety during the day are presented below (figure 9):
Figure 9: Safety during the day, Friedman’s Anova Test
The most preferred option is level 4, whereas level 1 and level 2 fall below the ranking of 4 (considered safe). As of note, the lower and upper quartile for level 4 are above the indication of 4, signifying that level 4 is perceived as the safest alternative during the day. Level 3 has various outliers which exemplifies the vast perception of this alternative‟s safety.
The box plots representing each alternative‟s safety at night are provided below (figure 10).
Figure 10: Safety during the night, Friedman’s Anova test
Though level 4 was considered as the safest alternative during the day, this is not true during the night time (figure 10). The vast amount of outliers in level 4 indicates the varying perceptions of safety. As for all other levels, their perception of safety remained more consistent. At night, both level 2 and 3 are perceived by majority to be relatively safe. As for level 1, the perception of safety varies in extremes as well, though majority perceived this alternative as neutral or very unsafe.
It is important to note that several outliers are present in level 4‟s box plot. In addition, level 2 and 3 have both lower and upper quartiles above an indication of 3, thus these two alternatives were considered neutral or safe.
The three figures (8,9,10) presented above have illustrated participants‟ perceptions of safety (both during the day and the night) as well as their preference for varying inclusions of vegetation. Though differences amongst the levels are evident, the Friedman‟s Anova test does not state where the largest difference occurs between the four visualizations. Thus, the Wilcoxon‟s test will be used.
The Wilcoxon‟s test is used in order to identify where the largest significance occurs between all four visualizations. This test was computed by using SPSS. Each visualization was compared to all others in order to find where the greatest significance occurred. The following are the results of the Wilcoxon test.
The test began by comparing level 1 to all other levels. As long as the significance remains below 0.05, then there is a significant difference between the two levels.
Table 4: Wilcoxon Test Greenery
For the statistical analysis, the two-tailed significance will be of importance. From level 1 to level 2 there is a significance of 0.034. Though this may seem as a great distinction between the two levels, there are greater significances between level 1 and levels 3 and 4 which include greater amounts of vegetation. The greatest significance occurred between level 1 and level 4, since the significance is 0.000. Alternatively, the significance between level 1 and level 3 is very great as well, due to the significance of 0.001.
Table 5: Wilcoxon Test Greenery
It is interesting that a greater significance occurred between level 2 and level 1 than occurred between levels 2 and levels 3. The greatest significance is evident between level 2 and level 4 with a significant value of 0.000.
Table 6 depicts the comparison of level 3 to all other visualizations.
Table 6: Wilcoxon Test Greenery
A great significance is found between level 3 and level 1, whereas the greatest significance occurs between level 1 and level 4, again at a significance of 0.000. It is important to note that nearly the same impact between level 1 and level 4 can be produced between level 1 and level 3. These options will be analyzed in the section of recommendations and conclusion.
Level 4 in comparison to all other visualizations:
Evidently, level 4 has a significant difference between all other levels. Thus, the greater inclusion of vegetation holds the greatest significant difference to all other levels. Though the Wilcoxon test has identified that the largest significant difference occurs between various levels and level 4, this non-parametric test does not indicate how strong the difference is between the two levels. In order to find this, the effect sizes and Cohen‟s D will be used.
4.1.4 Effect Sizes and Cohen’s D
Effect size depicts whether the scale of 1-5 was interpreted in the same manner by all participants. The results of this test depicted that as the greening increased, the significance of the options increased as well. The table below represents the meaning of value ranges for Cohen‟s D, and how they may be interpreted.
Table 8: Range of Cohen D values
|Range of Values||Meaning and Example|
|0 – 0.2||Small and Negligible (example of introducing a mole hill on a large mountainous area)|
|0.5||Moderate (example of adding bay windows on a building)|
|0.8||Large effect (example of adding a mountain in a field)|
Table 9: Effect Sizes and Cohen’s D
|Change from Control||Effect Size||Cohen’s D|
|G1 v. G2||0.2816937||0.5871651|
|G1 v. G3||0.3428446||0.7299288|
|G1 v. G4||0.7026401||1.9749691|
|D1 v. D2||0.1589393||0.3219715|
|D1 v. D3||0.2461728||0.5079782|
|D1 v. D4||0.4959718||1.1433467|
|N1 v. N2||0.2674364||0.5550918|
|N1 v. N3||0.3022788||0.6342271|
|N1 v. N4||0.5349835||1.2664393|
Once again the greatest significant difference occurs between the control level (1) and level 4. The power of the difference is shown by the Cohen‟s D value being 1.9749691. The immensity of this impact is repeated in relation to safety during both the day and the night. For the day, level 4 has a Cohen‟s D value of 1.1433467 in comparison to the control. This strength in significant difference remains strong even during the night between the control level and level 4, with a Cohen‟s D value of 1.2664393 Evidently the level with the greatest inclusion of greenery had the greatest impact in relation to all other levels, regarding vegetation, and safety in both night and day. A limitation to this test is that one does not know whether the impact is negative or positive, however only the strength of the difference. Qualitative data presents the perception (either negative or positive) of each level of greenery, as well as safety during day and night. In regards to effect sizes, as the level of greenery increases, so do the effect size values, which also holds true in relation to the Cohen‟s D values.
In addition to the quantitative tests described above, the survey included two open-ended question that gave the surveyee the opportunity to explain their ideas, and reasoning behind a specific ranking. This section will expose the qualitative replies of participants in order to broaden the understanding of the quantitative data results.
Having reviewed the responses of participants, there are a total of 32 people who mentioned vegetation as a positive element that enhanced their walking experience, and aided in their ranking. The following are citations from select participants:
Table 10: Qualitative Survey Responses
|Gender||Response||Positive or Negative connotation to vegetation|
|Female||The trees add onto the desirability of walking during the day.||Positive|
|Female||The trees makes it feel like the area is kept better||Positive|
|Female||The trees make me feel that the city has a long history||Positive|
|Male||The increased planter size will make the pedestrian feel trapped and claustrophobic||Negative|
|Male||Walking in the open makes it look safer than when its covered by trees||Negative|
|Male||I liked the green during the day and it didn‟t feel congested at night. Without the trees, the street lamps had a greater positive effect and provided more visibility||Positive|
|Male||Option 3 was the best, it has the most natural foliage while still looking bright at night||Positive|
|Female||Option 2 – I feel that this is the most people-friendly in terms of safety for children and women as long as people were instructed to not walk too close to the road way||Neutral|
Null hypothesis: Greenery will not have a significant impact on participants
Hypothesis: Greenery will have a significant impact on participants.
Based on the statistical analysis as discussed above, and the qualitative results, the null hypothesis is rejected and the hypothesis is accepted. Thus, greenery does have significant impact on the streetscape.