Overview of Data and Methodology
Summary of the Information Collected
This research was based on the creep test results corresponding to 118 FRC specimens, tested under sustained flexural loads for at least 90 days. For each of these cases, the information collected included parameters concerning the following aspects:
  Concrete matrix: maximum aggregate size, average compressive strength at 28 days (f_{c}, MPa).
  Fibres employed: material (glass, steel or synthetic), length, aspect ratio, and fibres content as volume fraction.
  Flexural residual strength: limit of proportionality f_{L}, and residual strength parameters f_{R1}, f_{R2}, f_{R3} and f_{R4} (MPa) as per the standard EN 14651 [6].
  Sustained load during the creep test typified by the residual capacity, or load ratio IFa [7].
  Creep coefficients at the ages of 14, 30, and 90 days [7].
  Crack opening rates (CORs) between 0 and 14 days, 14 and 30, and 30 and 90 days (COR^{014}, COR^{1430}, and COR^{3090} respectively) [7].
Representative values for each of the variables are given in Table 1 for reference. Minimum, average and maximum values were intended to be descriptive of the entire set of data and were calculated as the 5, 50, and 95 % percentiles. The percentage of missing values for each variable represents those cases for which complete information could not be obtained.
Table 1 Summary of the information in the database
Minimum (5 %percentile) 
Average 
Maximum (95 %percentile) 
Percentage of missing values (%) 

Fibres content, Vf (%) 
0.47 
0.58 
0.95 
0.0 
Fibre material 
n.a. 
n.a. 
n.a. 
0.0 
Fibre length (mm) 
35.00 
42.89 
54.00 
0.0 
Fibre slenderness 
45.00 
71.09 
158.00 
0.0 
Max. aggregate size (mm) 
10.00 
12.46 
20.00 
11.0 
Compr. strength fC (MPa) 
34.00 
42.31 
56.54 
0.0 
Limit of proportionality fL (MPa) 
2.97 
4.32 
6.04 
23.7 
Residual strength fRi (MPa) 
0.87 
3.65 
9.55 
0.8 
Residual strength fR2 (MPa) 
0.79 
3.92 
9.06 
36.4 
Residual strength fR3 (MPa) 
0.78 
4.01 
10.09 
32.2 
Residual strength f_{R4} (MPa) 
0.60 
3.80 
9.57 
31.3 
Load ratio IFa (%) 
36.90 
66.21 
89.56 
0.0 
Creep coefficient at 14 days 
0.09 
0.70 
1.93 
0.8 
Creep coefficient at 30 days 
0.11 
0.91 
2.19 
0.0 
Creep coefficient at 90 days 
0.31 
1.26 
3.11 
2.5 
COR between 014 days 
0.86 
15.11 
41.07 
0.0 
COR between 1430 days 
0.20 
3.58 
10.15 
0.0 
COR between 3090 days 
0.27 
1.51 
4.16 
0.0 
Treatment of Missing Values
As the sources of these data did not always report the mix characteristics, their flexural response or their flexural creep behaviour in the same way, some of the values of interest had not been reported.
This is quite common in the context of data mining applications or multivariate statistics [8], and different strategies can be adopted to moderate the impact of missing cases on the reliability of the information that can be extracted from a set of data. In this case, the prevalence of missing values was relatively small and did not justify complete elimination of any cases. Instead, a multiple imputation by fully conditional specification [9] was done to simulate the missing values taking advantage of the multiple correlations between the different variables.