Professional Documents
Culture Documents
Abstract. This study would examine the effect of technology on the processing ability of
messages, small and large parcels in Australia Post facilities for data comprehension purposes
and whether variables have any effect on each other. The information includes the date, year,
facility, letters, small parcels, and big parcels. 2 dependent groups, 2 independent groups and
more than 2 independent groups form the 3 primary fields to be investigated. Data for learning
changes over time will be plotted by bar graphs, and the Shapiro-Wilk test will validate
normality. This will allow us to determine whether to pick a parametric or non-parametric test to
search for statistical variations. These experiments will analyze major variations within each
dataset and the effects of data trends, patterns and findings will be addressed.
Keywords: independent groups, paired groups, p-value, testing hypothesis
1 Introduction
This report offers an analysis of the effect of technology on letters, small and large parcels in
Australia Post's processing capacity. Year, services, messages, small parcels and large parcels are
the data to be examined. Furthermore, data is broken down into 3 angles: 2 separate groups, 2
dependent groups and more than 2 separate groups. The reasons for the necessary collection, data
are gathered, normality tests are conducted and statistical tests are used to draw the conclusions.
Bar graphs are selected to display improvements over time and the Shapiro-Wilk test is selected
for normality validation. This in turn confirms whether parametric or non-parametric
experiments are conducted. Zero hypotheses and alternative hypotheses are described. This
discusses any major variations between classes. After various tests, inferences and reflections
will be identified.
2 Proving significant difference among 2 independent groups
Independent variable does not depend on remaining variables but relies on the remaining
variables. Year and region (Australia post-processing region: facility) will rely on this as they
decide independent variables. We can carry on with year and facility for 2 separate groups
because they don't rely on the remaining variables.
2.1 Plotting and explaining data
Tests of Normality
Kolmogorov-Smirnov Shapiro-Wilk
Statistic df Sig. Statistic df Sig.
Year .341 2190 .000 .637 2190 .000
Australia Post processing .223 2190 .000 .793 2190 .000
area
a. Lilliefors Significance Correction
Fig. 5. Bar map with a section for messages, tiny and (Tullamarine, Mascot, Brisbane)
Bar charts display improvements over time, and area results are shown in Figure 5 with letters,
large and small parcels while the y-axis averages[1]. In general , the results look similar,
particularly for all the letters.
4.2 Discussing and validating use of non-parametric test
Q-Q plots of the variables are executed to ascertain distribution nature of data.
Fig. 6. Quantile-Quantile plots for letters, small and large parcels
As seen in figure 6, the points are not a line roughly straight that shows that the quantities do not
derive from the normal distribution. We will run Shapiro Wilk data normality test[1]. The null
hypothesis is that the sample is naturally split, while the alternative hypothesis is that the sample
is not separated. According to Table 5, the p-value is below 0.05. Therefore null hypotheses may
be dismissed, and non-parametric tests can be employed as data are not distributed normally.
4.3 Discussion over results
Since the data is not distributed normally and represents three separate samples, Kruskal-Wallis
test can be performed [6]. The test is conducted to see if two or more samples come from the
same distribution. Null hypothesis indicates that the letters, small plots, and large parcels are
very different, and the alternative hypothesis notes that there is a difference[5]. Test findings in
Table 6 indicate that the p-value is smaller than 0.05 (0.00), meaning zero hypothesis is refused.
In conclusion, the three separate categories have a big difference: letters, small parcels, large
parcels.
Table 6: Kruskal-Wallis test for >2 independent samples
References
1. Dedecker, J. and Saulière, G.: The Mann–Whitney U-statistic for α-dependent sequenc-es.
Mathematical Methods of Statistics. 26(2), 111-133 (2017).
2. Hosseini, R. and Takemura, A.: An objective look at obtaining the plotting positions for QQ-
plots. Communications in Statistics-Theory and Methods. 45(16), 4716-4728 (2016).
3. Huang, L.F.: Adjusted Wilcoxon signed rank test tables for ratio of percen-tiles.
Communications in Statistics-Simulation and Computation. 46(7), 5763-5771 (2017).
4. Kelleher, C. and Wagener, T.: Ten Guidelines for Effective Data Visualisation in Scien-tific
Publications. In Environmental Modelling and Software, 822-827 (2011).
5. List, J.A., Shaikh, A.M. and Xu, Y.: Multiple hypothesis testing in experimental econom-ics.
Experimental Economics. 22(4), 773-793 (2019).
6. Richardson, J.T.: Kruskal–Wallis Test (2018).