Professional Documents
Culture Documents
Confidence Interval and Prediction Estimates - Statistical Hydrology
Confidence Interval and Prediction Estimates - Statistical Hydrology
Assigment No. 5
Topic Helsel et al. Chapter 3 - Confidence Interval and Prediction Estimates
We will use the snow dataset again, but use column K, 3-6 Forest for all the analyses (which is under the
evergreens).
Solution:
1) Summary statistics
(𝑋𝑖 − 𝑋)3 From lowest
2
Rank depth (cm) 𝑌𝑖 𝑋𝑖 − 𝑋 𝑋𝑖 − 𝑋 𝑑𝑖 𝑠3 to highest 𝑑𝑖
1 13.710 2.618 -2.290 5.244 1.930 -4.499 0.000
2 14.590 2.680 -1.410 1.988 1.050 -1.050 0.000
3 15.420 2.736 -0.580 0.336 0.220 -0.073 0.220
4 15.640 2.750 -0.360 0.130 0.000 -0.017 0.880
5 15.640 2.750 -0.360 0.130 0.000 -0.017 1.050
6 16.520 2.805 0.520 0.270 0.880 0.053 1.320
7 16.960 2.831 0.960 0.922 1.320 0.331 1.820
8 17.460 2.860 1.460 2.132 1.820 1.166 1.930
9 18.060 2.894 2.060 4.244 2.420 3.275 2.420
The summary statistics of the 3-6 Forest data provide a depth Mean of 16.00 cm, a Median of 15.64 cm,
Standard Deviation of 1.387 cm, Interquartile Range of 2.205 cm, the Skewness Coefficient and Quartile
Skew are -0.1337 and 0.424 respectively.
2) Boxplot
Using R
Probability plot
𝑖 19.000
𝑝=
Rank Depth (cm) (𝑛 + 1) Zp
1 13.710 0.100 -1.282 18.000
2 14.590 0.200 -0.842
Snow depth (cm)
The dataset is very symmetric, it is represented in the low skewness factor. Also, it is observed in the boxplot
and the probability plot, where the linear fit trendline matches the distrubution of the values.
3) Confidence interval and prediction estimates
The 96.09 % non-parametric confidence interval for the median is (14.59 cm, 16.96 cm). The 95 %
parametric confidence interval is (14.93 cm, 17.07 cm). These two intervals are close, however since the
data shows to be symmetric, the parametric approach is suggested.
The 95% parametric prediction interval for the median is (12.63 cm, 19.37 cm), and the 95% non-parametric
prediction interval is (13.71 cm, 18.06 cm). In this case the non-parametric approach shows a wider range
compared to the parametric approach.
4) Repeat the calculation of the nonparametric 95% confidence intervals by hand.
5) Based on parts 1 and 2, which is the most appropriate technique to estimate the intervals.
Based on the summary statistics (skewness factor) which is low, and from the plots that show a symmetric
distribution, the suggested approach to compute the confidence and prediction intervals is the parametric
one.
6) Based on part 3, if I have a snow depth in the field of 26.05 cm, does this indicate the field snow depth is
different than the forest?
The value of 26.05 cm of snow depth falls out of the confidence and prediction intervals, which means that
this one probably belongs to another group/area.