Appendix A – Analysis and statistics

Linear Interpolation of Results

During the analysis of the data, it was observed that the workload of the two groups would at some point cross between the 25mm and 50mm kerb upstand. Theoretically this point would be the location which both the Visually Impaired and Physically impaired user groups would have an equal level of workload.

Assuming the continuing trends of each group it can be mathematically calculated where that location could be using a simple Linear Interpolation to establish the mid-point using the Equation 1.

Equation 1: Linear Interpolation

𝑦 = 𝑦_1+(π‘₯βˆ’π‘₯_1)(𝑦_2βˆ’π‘¦_1)/(π‘₯_2 βˆ’π‘₯_1 )

Where:

  • π‘₯ =π‘₯_1+(π‘₯_2βˆ’π‘₯_1)/2
  • π‘₯_1 and 𝑦_1 are the first coordinates, and
  • π‘₯_2 and 𝑦_2 are the second coordinates
  • π‘₯ is the point to perform the interpolation
  • 𝑦 is the interpolated value

Note that the Low Vision and Moderate Vision groups’ data were combined to simplify the calculation.

This yielded the result in TableApp 1, below.

TableApp 1: Intersection Points
Intersection Full-Batters Half-Battered Bull-Nose Cycle segregation
x 44.50 47.50 42.50 40.25
y 28.75 29.00 23.75 23.50

The interpolation suggests that a kerb upstand in the range of 40mm-48mm would be of equal Medium workload for both Visual impaired and Physical impaired users’ groups. Considering engineering standards this translates to 40mm or 45mm upstands.

Further testing of different kerb upstand heights between 25mm and 50mm, with a focus on the 40 – 48mm range, could collect qualitative information on how users interact with kerbs in this range.

Cycle Segregation Kerb / Full Batter Kerb Discussion

The Cycle Segregation kerb used in the experiments was a 209 Cycle Segregation Unit from Marshalls.

It is acknowledged that this has a 45 degree splay which would be the same as the Full Batter kerb and that participants experience of the kerb profile face would therefore be similar.

As shown in TableApp 2, below, this similarity is broadly replicated in the results. The difference between the Full Batter and Cycle segregation kerbs from Low Vision and Moderate Vision participants at 25mm being within one workload point of each other. However, at 50mm a variation of 12 to 7 workload points is evident. Looking at the mentally and physically weighted results, TableApp 3 and TableApp 4 shows that both physical and mental attributes played a key part in the visually impaired user groups perceived workloads (as discussed in report section 3.2.2).

For the Physical user group at 50mm the workload was the same for both kerb types, but a difference of 5 workload points is seen at the lower 25mm upstand in favour of the Cycle segregation kerb. Assesment of the weighted scores for the Physical user group suggests this is predominated by the physical workload factor, as shown inTableApp 4. This is particularly evident on the Full Batter Kerb at 50mm where a 10 point increase on physically weighted workload can be observed.

The overall width of the Cycle segregation kerb is 290mm, which is over twice as large as the width of the Full Batter Kerb at 125mm. This suggests that the width of the kerb, in addition to the profile played a significant part in the perceived physical and mental workloads experienced by the users of the kerbs and therefore they were observed and traversed differently to the Full Batter kerb despite the similarities.

TableApp 2: Unweighted Perceived Workload CS vs FB
Kerb PHYSICAL LOW VISION MODERATE VISION BASELINE
FB25 14 42 31 7
FB50 33 29 24 7
CS25 9 41 30 5
CS50 33 17 17 9
TableApp 3: Mental Weighted Perceived Workload CS vs FB
Kerb PHYSICAL LOW VISION MODERATE VISION BASELINE
FB25 8 37 33 13
FB50 12 24 25 10
CS25 8 38 30 8
CS50 32 14 17 13
TableApp 4: Physically Weighted Perceived Workload CS vs FB
Kerb PHYSICAL LOW VISION MODERATE VISION BASELINE
FB25 16 27 18 8
FB50 43 27 18 8
CS25 12 36 24 5
CS50 36 13 12 10

Analysis: Inferential Statistics for TLX findings

NASA-TLX non-parametric Wilcoxon Test

The factors and conditions of the experiment are given in TableApp 5, below.

  • Factor 1 is Kerb Type, with four levels.
  • Factor 2 is Kerb Height, with five levels.

These constitute the Independent Variables. The Dependent (measured) Variables were the TLX rating scores given by the participants after each trial. The test used was the non-parametric equivalent of the t-test.

TableApp 5: Operationalised kerb size and type table.
Type 20 25 50 60 100
Type 1 - Full-Batter - SP No Information FULLBAT_25 FULLBAT_50 FULLBAT_60 - No Information
Type 2 - Half-Batter - HB2 No Information HALFBAT_25 HALFBAT_50 HALFBAT_60 HALFBAT_100
Type 3 - Bull-Nose - BN No Information BULLNOSE_25 BULLNOSE_50 JBULLNOSE_60 BULLNOSE_100
Type 4 - Cycle segregation CYCLESEGREGATION_20 CYCLESEGREGATION_25 CYCLESEGREGATION_50 No Information No Information

Wilcoxon signed-rank tests were conducted to identify whether the differences in outcomes between kerb types and heights used in the rig experiment were statistically significant. In this study, the measurements are based on the TLX rating scale from low (0%) and high (100%) for mental demand, physical demand, temporal demand, effort and frustration. While the scale for the performance is from poor (0%) to good (100%).

We only present the outcomes that were statistically significant in the TableApp 6 and TableApp 7 below. Significance indicates where the probability of the result is less than p < 0.05 (5%) indicated by an asterisk (*). Non-significant results indicate that the result may be due to chance, not that there was no difference. Tests were carried out as 2-tailed, meaning that the direction of difference was not predicted. Any missing data led to the specific comparison being excluded. The full test outputs are in the A.3.2 Wilcoxon Signed Ranks Test, below.

TableApp 6: Wilcoxon Signed Ranks Test Results-1
Group HB50_Mean - HB25_Mean FB50_Mean - HB25_Mean BN50_Mean - HB25_Mean CS25_Mean - HB50_Mean FB50_Mean - HB50_Mean HB60_Mean - HB100_Mean
Low Vision (5) 0.042* Non-significant 0.043* 0.043* Non-significant 0.043*
Moderate Vision (4) Non-significant Non-significant Non-significant Non-significant Non-significant Non-significant
Physical Group (6) Non-significant 0.028* 0.027* 0.046* 0.028* Non-significant

Visual groups

While considering the Low Vision group, test between Half-Batter 50 and 25 (HB50_Mean - HB25_Mean) there is statistical significance difference between the mean HB50 and HB25 at a greater 95% level of confidence (p-value=0.042).

Still on the low vision, similar outcomes between the following: BN50_Mean x HB25_Mean, CS25_Mean - HB50_Mean, HB60_Mean x HB100_Mean, CS50_Mean x HB100_Mean (See Tables 3 and 4 above).

TableApp 7: Wilcoxon Signed Ranks Test Results-2
Group CS25_Mean - HB100_Mean CS20_Mean - HB100_Mean CS50_Mean - HB100_Mean CS_Mean - HB_Mean CS_Mean - FB_Mean CS_Mean - BN_Mean
Low Vision Non-significant Non-significant 0.043* Non-significant Non-significant Non-significant
Moderate Vision Non-significant Non-significant Non-significant Non-significant Non-significant Non-significant
Physical Group 0.046* 0.046* Non-significant 0.028* 0.028* 0.028*

From both the qualitative and quantitative results, we can conclude that there is evidence within the Low Vision group for significant differences in the perception of workload between heights 25mm and 50mm, primarily as a result of mental workload. This is likely due to the difficulty in detectability of the lower kerbs below 50mm with canes, dog, or feet. The 60mm -100mm difference is a response to increasing difficulty at 100mm.

The lack of differences between Low vision and Moderate vision is likely attributable to the overall similarity of scores between these two groups. These differences and too small to be detected at the test power and sensitivity.

Physical Group

Similarly, for the Physical group, there is evidence of statistical significance of differences in Full-Batter 50 and Half-Batter 25 (FB50_Mean x HB25_Mean) with the p-value equal 0.028 (the confidence level greater than 99%). similar outcomes between the following: BN50_Mean x HB25_Mean, CS25_Mean x HB50_Mean, FB50_Mean x HB50_Mean, CS25_Mean x HB100_Mean, and CS20_Mean x HB100_Mean (See TableApp 6 and TableApp 7 above).

From both the qualitative and quantitative results we can conclude from this that there is evidence within the Physical group for significant differences in the perception of workload between heights 25mm and 50mm, primarily as a result of physical workload. There is also evidence for differences between the Full-Batter and Half-Batter kerbs at 50mm upstand, attributable to differences in kerb profile and its interaction with wheelchairs, canes, and feet. to differences in physical kerb profile and its interaction with wheelchairs, sticks, and feet.

Differences between Kerb Types

We looked at the difference between kerb types without a specific kerb height. The outcomes only showed statistically significant differences for the Physical group.

For the Physical group, there was evidence of statistical significance of differences between

  • Cycle Segregation and Half-Batter (CS_Mean x HB_Mean);
  • Cycle Segregation and Full-Batter (CS_Mean x FB_Mean); and
  • Cycle Segregation and Bull-Nose (CS_Mean x BN_Mean),

with each having the p-value of <0.05* significance level (See TableApp 7 above).

Wilcoxon Signed Ranks Test

Wilcoxon signed-rank tests were conducted to identify whether the differences in outcomes between kerb types and heights used in the rig experiment are statistically significant. In this study, the measurements are based on the TLX rating scale from low (0%) and high (100%) for mental demand, physical demand, temporal demand, effort and frustration. While the scale for the performance is from poor (0%) to good (100%).