STA680 LAB SPSS Week 8
Data Description
Variable Description Unit
Information skills level (X1) Score of Information No unit
processing skills
Comprehension level (X2)
Data Exploration
- Data Visualization (dependent variable)
- Data Descriptive (independent variables)
CHAPTER 4
RESULTS AND DISCUSSION
4.0 Introduction
4.1 Data visualization of age groups of 174 children
Figure displays distribution of age groups of 174 children. Majority of children belongs to the
group 3 (37.93%), followed by group 2 (31.62%), group 1 (25.88%) and group 0 (4.60%)
respectively…
Figure 1. The bar graph age groups of 174 children
Figure 2 depicts the distribution of age groups of 174 children. It is observed that majority of
the children belong to group 3 (37.93%), followed by group 2, group 1 and group 0
respectively.
Figure 2. The pie chart of age groups of 174 children
4.2 Data descriptive of independent variables
Table 1 provides the descriptive statistics of information levels, comprehension level and
arithmetic ability of 174 children. Across all variables, comprehension level (9.99) has the
highest mean followed by information level (9.47) and arithmetic level (8.98) respectively.
Table 1. Descriptive Statistic of Independent Variables
Variable Mean Median Std. Range Skewness Kurtosis
Deviation Statistic Std. Statistic Std.
Error Error
Information 9.47 10 2.900 16 0.092 0.184 0.022 0.366
Comprehension 9.99 10 2.973 18 0.092 0.184 0.396 0.366
Arithmetic 8.98 9 2.302 12 0.412 0.184 -0.079 0.366
4.3 Data pre-processing
4.3.1 Determination of Missing Data
Table 2 shows that there are no missing data across the three variables belongings of 174
children.
Table 2. Data Missingness Detection
Variable Valid Cases Missing Total
N Percent N Percent N Percent
Information 174 100.0% 0 0.0% 174 100.0%
Comprehension 174 100.0% 0 0.0% 174 100.0%
Arithmetic 174 100.0% 0 0.0% 174 100.0%
4.3.2 Detection of Univariate Outlier
Table 3 provides the summary of Q-Q plots and boxplots of all independent variables. Across
arithmetic level (X1), there are possible outliers as on the higher sides of values from both
diagrams. Across Q-Q plots, 4 values seem to be deviating from the straight line. Meanwhile,
from the box plots, several extreme values are observed on the higher tail of boxplot.
(continue interpret for X2 and X3)
Table 3. Summary of Q-Q plots and boxplots of all independent variables.
The Q-Q plot for Information Level The box plot for Information Level
The box plot for Comprehension Level
The Q-Q plot for Comprehension Level
The box plot for Arithmetic Ability
The Q-Q plot for Arithmetic Ability
4.3.2.1 Treatment of Univariate Outlier
Table 4 provides the summary of univariate outliers across three independent variables.
Across information level, score belonging to child 136 with score of 19 has z value equivalent
to 3.29 9more than +3). Therefore, 95th percentile of 14 was used to replace the original value.
…………..
Table 4. Summary of univariate outliers across three independent variables
Variable Low Extreme Value High Extreme Value Treatment
Information - Child 136 (19) Child 136 (14)
Arithmetic
Statistical Test for a Multivariate Normality Assumption Coefficient
1. Specify 𝑯𝟎 𝒂𝒏𝒅 𝑯𝟏
𝐻𝑜 : 𝐷𝑎𝑡𝑎 𝑓𝑜𝑙𝑙𝑜𝑤 𝑎 𝑡𝑟𝑖𝑣𝑎𝑟𝑖𝑎𝑡𝑒 𝑛𝑜𝑟𝑚𝑎𝑙 𝑑𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑖𝑜𝑛
𝐻1 : 𝐷𝑎𝑡𝑎 𝑑𝑜𝑒𝑠 𝑛𝑜𝑡 𝑓𝑜𝑙𝑙𝑜𝑤 𝑎 𝑡𝑟𝑖𝑣𝑎𝑟𝑖𝑎𝑡𝑒 𝑛𝑜𝑟𝑚𝑎𝑙 𝑑𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑖𝑜𝑛
2. Set significance level (𝜶 = 𝟎. 𝟎𝟓), find critical chi-squared value at v=p
Using 𝜒(𝑣 = 3) 𝑎𝑡 (𝛼 = 0.05 = 2.366)
Reject 𝐻0 𝑜𝑓 % 𝑜𝑓 𝑚𝑎ℎ𝑎𝑙𝑎𝑛𝑜𝑏𝑖𝑠 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 2.366 𝑖𝑠 𝑎𝑡 𝑙𝑒𝑎𝑠𝑡 50%
(P (mahalanobis distance < 2.366) > 0.5)
3. Calculate Mahalanobis Distance for each data and compare with critical value
𝟗𝟏
4. Altogether there are 𝟏𝟕𝟒 % = 53%
Since Mahalanobis Distance is 53% more than 50%, we fail to reject 𝐻0
5. We can conclude that the data is trivariate normally distributed
4.4 MANOVA Application
4.4.1 Descriptive of multiple variables across 4 age group of 174 children
The summary of group means vectors:
Let 𝑥̅1 = the mean of group 1 based on X1, X2 and X3
9.63
= (10.63)
9.00
Let 𝑥̅2 = the mean of group 2 based on X1, X2 and X3
8.73
= (9.36)
8.96
Let 𝑥̅3 = the mean of group 3 based on X1, X2 and X3
10.18
= (10.45)
8.84
Let 𝑥̅4 = the mean of group 4 based on X1, X2 and X3
9.36
= (9.97)
9.12
4.4.2 Testing the equality of variance covariance matrices across 4 different age
groups.
𝐻0 : 𝐴𝑙𝑙 𝑜𝑓 𝑣𝑎𝑟𝑖𝑎𝑛𝑐𝑒 𝑐𝑜𝑣𝑎𝑡𝑖𝑎𝑛𝑐𝑒 𝑚𝑎𝑡𝑟𝑖𝑐𝑒𝑠 𝑎𝑟𝑒 𝑡ℎ𝑒 𝑠𝑎𝑚𝑒 𝑎𝑐𝑟𝑜𝑠𝑠 4 𝑎𝑔𝑒 𝑔𝑟𝑜𝑢𝑝𝑠
𝐻1 : 𝐴𝑡𝑙𝑒𝑎𝑠𝑡 𝑜𝑛𝑒 𝑝𝑎𝑖𝑟 𝑜𝑓 𝑎 𝑔𝑟𝑜𝑢𝑝 𝑣𝑎𝑟𝑖𝑎𝑛𝑐𝑒 𝑐𝑜𝑣𝑎𝑡𝑖𝑎𝑛𝑐𝑒 𝑚𝑎𝑡𝑟𝑖𝑐𝑒𝑠 𝑎𝑟𝑒 𝑑𝑖𝑓𝑓𝑒𝑟
Reject 𝜌 < 𝛼 = 0.05
Since 𝜌 = 0.064 > 0.05, 𝑑𝑜 𝑛𝑜𝑡 𝑟𝑒𝑗𝑒𝑐𝑡 𝐻0
At 𝛼 = 0.05, all of variance covariance matrices are the same across 4 age group. Therefore,
MANOVA application can proceed on this multivariate data.
4.4.2.3 Testing of significant differences between the four group mean vectors based
on three variables
𝐻0 : 𝜇1 = 𝜇2 = 𝜇3 = 𝜇4
𝐻1 : 𝐴𝑡𝑙𝑒𝑎𝑠𝑡 𝑜𝑛𝑒 𝑝𝑎𝑖𝑟 𝑜𝑓 𝜇𝑖𝑗 𝑑𝑖𝑓𝑓𝑒𝑟
𝛼 = 0.05
𝜌 = 0.219
Reject 𝜌 < 𝛼 = 0.05
Since 𝜌 = 0.064 > 0.05, 𝑑𝑜 𝑛𝑜𝑡 𝑟𝑒𝑗𝑒𝑐𝑡 𝐻0
At 𝛼 = 0.05, we can conclude that
*No need to do Bonferroni test if we fail to reject 𝐻0
4.4.2.4 Profile Plots
There are 3 patterns to check.
1. Are the profiles parallel?
Test for parallel profiles:
𝐻0 : 𝜇1𝑖 − 𝜇1,𝑖−1 = 𝜇2𝑖 − 𝜇2,𝑖−1 , 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖 = 2, … . , 𝑝
𝐻1 : 𝐴𝑡𝑙𝑒𝑎𝑠𝑡 𝑜𝑛𝑒 𝑝𝑎𝑖𝑟 𝑜𝑓 𝜇𝑖𝑗 𝑑𝑖𝑓𝑓𝑒𝑟
2. Are the profiles coincident?
Test for coincident profiles:
𝐻0 : 𝜇1𝑖 = 𝜇2𝑖 , 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖 = 1, … . , 𝑝
𝐻1 : 𝐴𝑡𝑙𝑒𝑎𝑠𝑡 𝑜𝑛𝑒 𝑝𝑎𝑖𝑟 𝑜𝑓 𝜇𝑖𝑗 𝑑𝑖𝑓𝑓𝑒𝑟
3. Are the profiles level (flat)?
Test for level (flat) profiles:
𝐻0 : 𝜇1𝑖 = 𝜇2𝑖 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖 = 1, … . , 𝑝
𝐻1 : 𝐴𝑡𝑙𝑒𝑎𝑠𝑡 𝑜𝑛𝑒 𝑝𝑎𝑖𝑟 𝑜𝑓 𝜇𝑖𝑗 𝑑𝑖𝑓𝑓𝑒𝑟
