Uploaded by patryk langer

Output for the paper

advertisement
Descriptives
All economies included
Conclusion and Info:
Mean (Average): The mean is the average value of a dataset. It
is calculated
by adding up all the values and then dividing
by the number of values. In
your data, the mean gives the
average score or level for each variable across
your
observations (such as countries or regions).
Variance: Variance measures how spread out the values in your
dataset are.
A high variance indicates that the numbers are
more spread out from the mean
and from each other, while a
low variance indicates that the values are closer
to the
mean and to each other.
For example, a high variance in "ICT skills
in the education
system" suggests that the levels of these skills vary
significantly across different observations.
Skewness: Skewness measures the degree of asymmetry of the
distribution of
values. A skewness close to zero indicates a
symmetrical distribution.
A positive
skewness means the tail on the right side of the
distribution is longer or
fatter, suggesting a concentration
of values at the lower end.Conversely,
a negative skewness
means the tail on the left side is longer or fatter,
indicating
a concentration of values at the higher end.
Kurtosis: Kurtosis measures the 'tailedness' of the
distribution. A normal
distribution has a kurtosis of about
0. Positive kurtosis (leptokurtic) indicates
a distribution
with heavier tails and possibly a sharper peak, suggesting
outliers are more likely.
Negative kurtosis (platykurtic) indicates a distribution
with lighter tails and a flatter peak, suggesting outliers are
less likely.
Standard Error (for Skewness and Kurtosis): This is a measure
of the uncertainty
or variability of the skewness and
kurtosis estimates. A larger standard error
Detailed Info:
ICT Regulatory Environment (Mean: 91.36, Variance: 24.83.
Skewness (-0.758):
Left-skewed distribution. This suggests
that most countries have a favorable
ICT regulatory
environment, with a few having significantly lower scores.
Kurtosis (1.655): Leptokurtic distribution. The distribution
has more outliers
than a normal distribution, indicating
some extreme values in the data. Regulatory
Quality (Mean:
69.55, Variance: 203.01). Skewness (-0.023): Near symmetrical
distribution. Most countries have similar levels of regulatory
quality. Kurtosis
(-1.012): Platykurtic distribution. The
distribution is flatter than a normal
distribution, with
fewer outliers. AI Scientific Publications (Mean: 11.05,
Variance: 118.41). Skewness (1.155): Right-skewed
distribution. Most countries
have a low number of AI
publications, with a few countries having significantly
higher numbers. Kurtosis (0.421): Slightly leptokurtic. The
distribution has
a few more outliers than a normal
distribution. Robot Density (Mean: 13.92,
Variance: 248.61).
Skewness (1.006): Right-skewed. Indicates that robot density
is generally low, but some countries have very high densities.
Kurtosis (0.101):
Close to mesokurtic. The distribution is
similar to a normal distribution
in terms of outliers. ICT
Skills in the Education System (Mean: 51.05, Variance:
674.60). Skewness (-0.817): Left-skewed. Suggests that most
countries have
relatively higher levels of ICT skills in
education. Kurtosis (0.061): Near
mesokurtic. The
distribution is similar to a normal distribution in terms
of
outliers. Tertiary Enrollment (Mean: 48.05, Variance: 233.58).
Skewness
(0.785): Right-skewed. Most countries have lower
tertiary enrollment rates,
with a few having very high
rates. Kurtosis (3.409): Leptokurtic. The distribution
has
more outliers, indicating some countries with exceptionally
high or low
enrollment rates. AI Talent Concentration (Mean:
15.78, Variance: 204.75).
Skewness (0.921): Right-skewed.
Indicates that AI talent is concentrated in
a few countries.
Kurtosis (1.264): Leptokurtic. There are more outliers than
in a normal distribution, pointing to countries with extremely
high or low
AI talent concentration. Mobile Apps Development
(Mean: 72.94, Variance: 73.65).
Skewness (-0.086): Near
symmetrical distribution. Kurtosis (3.805): Highly
leptokurtic. The distribution has a significant number of
outliers, suggesting
a few countries are far ahead in mobile
app development.
Descriptive Statistics
N
Statistic
Mean
Statistic
Std.
Deviation
Statistic
Variance
Statistic
Skewness
Statistic Std. Error
ICT regulatory
environment
Regulatory quality
AI scientific publications
Robot density
ICT skills in the
education system
Tertiary enrollment
AI talent concentration
Mobile apps
development
Adoption of emerging
technologies
Regulation of emerging
technologies
Government promotion
of investment in
emerging technologies
Investment in emerging
technologies
Computer software
spending
GERD financed by
business enterprise
Annual investment in
telecommunication
services
GERD performed by
business enterprise
R&D expenditure by
governments and higher
education
Knowledge intensive
employment
High-tech and mediumhigh-tech manufacturing
High-tech exports
PCT patent applications
ICT services exports
Valid N (listwise)
36
91.3631
4.98337
24.834
-.758
.393
36
36
36
36
69.5528
11.0517
13.9208
51.0508
14.24815
10.88150
15.76732
25.97301
203.010
118.407
248.608
674.597
-.023
1.155
1.006
-.817
.393
.393
.393
.393
36
36
36
48.0542
15.7783
72.9419
15.28317
14.30923
8.58179
233.575
204.754
73.647
.785
.921
-.086
.393
.393
.393
36
58.1358
24.44564
597.589
-.446
.393
36
59.0053
21.98380
483.288
-.458
.393
36
43.1997
22.98584
528.349
-.033
.393
36
48.6944
22.22982
494.165
.428
.393
36
33.9233
23.54880
554.546
.121
.393
36
53.2222
21.99193
483.645
-1.054
.393
36
77.9242
14.68832
215.747
-4.406
.393
36
22.7681
18.41139
338.979
.784
.393
36
25.2661
17.57265
308.798
.604
.393
36
59.7994
17.03941
290.342
-.128
.393
36
41.5817
21.51191
462.762
-.059
.393
36
36
36
36
20.6031
27.3758
35.0803
13.66158
25.14645
25.16845
186.639
632.344
633.451
.750
1.024
1.469
.393
.393
.393
Descriptive Statistics
Kurtosis
ICT regulatory
environment
Regulatory quality
AI scientific publications
Robot density
ICT skills in the
education system
Tertiary enrollment
AI talent concentration
Mobile apps
development
Adoption of emerging
technologies
Regulation of emerging
technologies
Government promotion
of investment in
emerging technologies
Investment in emerging
technologies
Computer software
spending
GERD financed by
business enterprise
Annual investment in
telecommunication
services
GERD performed by
business enterprise
R&D expenditure by
governments and higher
education
Knowledge intensive
employment
High-tech and mediumhigh-tech manufacturing
High-tech exports
PCT patent applications
ICT services exports
Valid N (listwise)
Statistic Std. Error
1.655
.768
-1.012
.421
.101
.061
.768
.768
.768
.768
3.409
1.264
3.805
.768
.768
.768
.243
.768
.155
.768
.005
.768
-.896
.768
-1.690
.768
.511
.768
23.719
.768
-.406
.768
-.631
.768
.007
.768
-1.000
.768
.559
-.016
1.711
.768
.768
.768
Factor Analysis
Correlation Matrix
Correlation Regulatory quality
AI scientific publications
Robot density
ICT skills in the
education system
Mobile apps
development
Regulation of emerging
technologies
Investment in emerging
technologies
Computer software
spending
GERD financed by
business enterprise
Annual investment in
telecommunication
services
GERD performed by
business enterprise
R&D expenditure by
governments and higher
education
Knowledge intensive
employment
High-tech exports
PCT patent applications
ICT services exports
Regulatory
quality
1.000
.044
.558
.595
AI scientific
publications
.044
1.000
.543
.311
Robot
density
.558
.543
1.000
.429
.344
-.022
.077
.830
.192
.528
.877
.243
.682
.289
.716
.556
.507
.434
.574
.124
.520
.427
.641
.418
.866
.685
.424
.858
.827
.105
.502
.549
.778
.175
.052
.265
-.206
.283
.689
-.150
Correlation Matrix
Correlation Regulatory quality
AI scientific publications
ICT skills in
Regulation of
the education Mobile apps
emerging
system
development technologies
.595
.344
.830
.311
-.022
.192
Robot density
ICT skills in the
education system
Mobile apps
development
Regulation of emerging
technologies
Investment in emerging
technologies
Computer software
spending
GERD financed by
business enterprise
Annual investment in
telecommunication
services
GERD performed by
business enterprise
R&D expenditure by
governments and higher
education
Knowledge intensive
employment
High-tech exports
PCT patent applications
ICT services exports
.429
1.000
.077
.375
.528
.664
.375
1.000
.249
.664
.249
1.000
.694
.326
.851
.518
-.013
.444
.508
.232
.630
.156
-.064
.019
.575
.149
.661
.577
.189
.682
.633
.381
.820
.425
.603
.360
.302
.310
.538
.547
.756
-.003
Investment in
emerging
technologies
.877
.243
.682
.694
Computer
software
spending
.289
.716
.556
.518
GERD
financed by
business
enterprise
.507
.434
.574
.508
.326
-.013
.232
.851
.444
.630
1.000
.506
.635
.506
1.000
.633
Correlation Matrix
Correlation Regulatory quality
AI scientific publications
Robot density
ICT skills in the
education system
Mobile apps
development
Regulation of emerging
technologies
Investment in emerging
technologies
Computer software
spending
GERD financed by
business enterprise
Annual investment in
telecommunication
services
GERD performed by
business enterprise
R&D expenditure by
governments and higher
education
Knowledge intensive
employment
High-tech exports
PCT patent applications
ICT services exports
.635
.633
1.000
.153
.336
.113
.750
.612
.672
.755
.588
.649
.855
.356
.660
.524
.876
.196
.287
.539
.048
.598
.628
.020
Annual
GERD
investment in performed by
telecommunic
business
ation services
enterprise
.124
.641
.520
.418
.427
.866
.156
.575
R&D
expenditure by
governments
and higher
education
.685
.424
.858
.577
Correlation Matrix
Correlation Regulatory quality
AI scientific publications
Robot density
ICT skills in the
education system
Mobile apps
development
Regulation of emerging
technologies
Investment in emerging
technologies
Computer software
spending
GERD financed by
business enterprise
Annual investment in
telecommunication
services
GERD performed by
business enterprise
R&D expenditure by
-.064
.149
.189
.019
.661
.682
.153
.750
.755
.336
.612
.588
.113
.672
.649
1.000
.372
.391
.372
1.000
.986
.391
.986
1.000
governments and higher
education
Knowledge intensive
employment
High-tech exports
PCT patent applications
ICT services exports
.035
.619
.639
-.327
.101
.104
.387
.776
.032
.372
.780
.012
Knowledge
intensive
employment
.827
.105
.502
.633
High-tech
exports
.549
.052
.283
.425
PCT patent
applications
.778
.265
.689
.603
.381
.302
.310
.820
.547
.756
.855
.524
.876
.356
.287
.539
.660
.598
.628
.035
-.327
.101
.619
.387
.776
.639
.372
.780
1.000
.580
.761
.580
.761
.129
1.000
.427
.132
.427
1.000
.182
Correlation Matrix
Correlation Regulatory quality
AI scientific publications
Robot density
ICT skills in the
education system
Mobile apps
development
Regulation of emerging
technologies
Investment in emerging
technologies
Computer software
spending
GERD financed by
business enterprise
Annual investment in
telecommunication
services
GERD performed by
business enterprise
R&D expenditure by
governments and higher
education
Knowledge intensive
employment
High-tech exports
PCT patent applications
ICT services exports
Correlation Matrix
ICT services
Correlation Regulatory quality
AI scientific publications
Robot density
ICT skills in the
education system
Mobile apps
development
Regulation of emerging
technologies
Investment in emerging
technologies
Computer software
spending
GERD financed by
business enterprise
Annual investment in
telecommunication
services
GERD performed by
business enterprise
R&D expenditure by
governments and higher
education
Knowledge intensive
employment
High-tech exports
PCT patent applications
ICT services exports
exports
.175
-.206
-.150
.360
.538
-.003
.196
.048
.020
.104
.032
.012
.129
.132
.182
1.000
Info On the Correlation matrix and exlploratory analysis: Regulatory
Quality: Highly correlated with "Investment in emerging
technologies"
(0.877) and "Regulation of emerging
technologies" (0.830).
This
suggests that better regulatory quality is often
associated with higher investment
and regulation in emerging
technologies. Moderate positive correlation with
"ICT skills
in the education system" (0.595) and "Robot density"
(0.558).
AI Scientific Publications: Strongly correlated with "Computer
software
spending" (0.716). This could indicate that regions
with higher computer
software spending also contribute more
to AI scientific publications.
Moderate
positive correlation with "Annual investment in
telecommunication services"
(0.520) and "GERD financed by
business enterprise" (0.434).
Robot Density: Very strong positive correlation with "GERD
performed
by business enterprise" (0.866) and "R&D
expenditure by governments
and higher education" (0.858).
High correlation with "Investment
in emerging technologies"
(0.682).
Mobile Apps Development: Shows a moderate positive correlation
with "ICT
skills in the education system" (0.375) and "ICT
services exports"
(0.538).
Investment in Emerging Technologies: Highly correlated with
"Regulatory
quality" (0.877), "Regulation of emerging
technologies" (0.851),
and "Knowledge intensive employment"
(0.855).
Indicates that investment
in emerging technologies is
closely linked with regulatory aspects and knowledge-intensive
employment.
GERD performed by Business Enterprise: Shows an almost perfect
positive correlation
with "R&D expenditure by governments
and higher education" (0.986),
suggesting these two
variables move almost in tandem.
High-Tech Exports: Displays a weak to moderate correlation
with most variables,
the strongest being with "Knowledge
intensive employment" (0.580).
PCT Patent Applications: Strongly correlated with "Investment
in emerging
technologies" (0.876) and "R&D expenditure by
governments and
higher education" (0.780).
Negative Correlations: Few variables, such as "ICT services
exports",
show negative correlations with others, but these
correlations are generally
weak.
For example, it's slightly negatively correlated with "AI
scientific
publications" (-0.206) and "Robot density" (0.150).
ICT Skills in the Education System: Shows a strong positive
correlation with
"Regulation of emerging technologies"
(0.664) and "Investment
in emerging technologies" (0.694).
This implies that regions with better
ICT skills in the
education system tend to have more robust regulation and
higher investment in emerging technologies.
Also, it has a moderate correlation
with "GERD performed by
business enterprise" (0.575) and "R&D
expenditure by
governments and higher education" (0.577), suggesting
a link
between ICT education and research and development activities.
Regulation of Emerging Technologies: Highly correlated with
"Investment
in emerging technologies" (0.851) and "Knowledge
intensive employment"
(0.820).
This indicates a strong association between the regulation of
emerging
technologies, investment in these areas, and
employment in knowledge-intensive
sectors.
Also shows a very strong correlation with "Regulatory quality"
(0.830), suggesting that high-quality regulation is often
accompanied by specific
regulations for emerging
technologies.
Computer Software Spending: Exhibits a strong correlation with
"AI scientific
publications" (0.716) and a moderate
correlation with "GERD performed
by business enterprise"
(0.612) and "R&D expenditure by governments
and higher
education" (0.588).
This suggests that areas with higher computer software
spending also have higher activity in scientific research and
R&D.
Annual Investment in Telecommunication Services: Shows
moderate positive correlation
with "AI scientific
publications" (0.520) but a relatively weak
correlation with
most other variables,
indicating that while there is some
relationship with
scientific publications, investment in telecommunications
does not strongly correlate with other areas like regulatory
quality or robot
density.
Knowledge Intensive Employment: Strongly correlated with
"Investment
in emerging technologies" (0.855) and
"Regulation of emerging technologies"
(0.820).
This implies that regions with higher employment in knowledgeintensive
sectors are also more likely to invest in and
regulate emerging technologies.
Also shows a very strong correlation with "Regulatory quality"
(0.827),
indicating that higher regulatory quality is often
associated with more knowledge-intensive
employment.
High-Tech Exports: Shows a moderate positive correlation with
"Knowledge
intensive employment" (0.580). This suggests that
regions with more high-tech
exports also tend to have higher
employment in knowledge-intensive sectors.
Relatively weak
correlations with other variables like "AI
scientificpublications" (0.052) and "Robot density" (0.283).
ICT Services Exports: Shows a moderate positive correlation
with "Mobile
apps development" (0.538). This could indicate
that regions with a higher
level of mobile app development
also have higher exports in ICT services.
Generally, shows weak correlations with other variables,
including slightly
negative correlations with a few, though
none are strongly negative
KMO and Bartlett's Test
Kaiser-Meyer-Olkin Measure of Sampling
Adequacy.
Bartlett's Test of
Approx. Chi-Square
Sphericity
df
Sig.
.778
581.855
120
<.001
Kaiser-Meyer-Olkin (KMO) Measure of Sampling Adequacy:
The KMO statistic is a measure of how well the variables in
your dataset are
related. It ranges from 0 to 1, with values
closer to 1 indicating that the
variables have more in
common, making your dataset more suitable for factor
analysis.
A KMO value of 0.778 suggests that your dataset is adequately
suitable for
factor analysis. Generally, a KMO value greater
than 0.6 is considered acceptable.
A value closer to 0.8 or above is considered good, indicating
that the patterns
of correlations are relatively compact and
so factor analysis should yield
distinct and reliable
factors.
Bartlett's Test of Sphericity:
This test checks the null hypothesis that the correlation
matrix is an identity
matrix, which would indicate that the
variables are unrelated and therefore
unsuitable for factor
analysis.
The test yields an approximate Chi-Square value of 581.855
with 120 degrees
of freedom and a significance level of less
than 0.001.
This very low significance
level (Sig. < .001) leads to
rejecting the null hypothesis, meaning that
the correlation
matrix is not an identity matrix.
The rejection of the null hypothesis in Bartlett's Test
suggests that there
are significant relationships among the
variables, and thus factor analysis
is appropriate.
Communalities
Regulatory quality
AI scientific publications
Robot density
ICT skills in the
education system
Mobile apps
development
Initial
1.000
1.000
1.000
1.000
Extractio
n
.884
.856
.831
.685
1.000
.668
Regulation of emerging
1.000
technologies
Investment in emerging
1.000
technologies
Computer software
1.000
spending
GERD financed by
1.000
business enterprise
Annual investment in
1.000
telecommunication
services
GERD performed by
1.000
business enterprise
R&D expenditure by
1.000
governments and higher
education
Knowledge intensive
1.000
employment
High-tech exports
1.000
PCT patent applications
1.000
ICT services exports
1.000
Extraction Method: Principal Component
Analysis.
.827
.907
.835
.777
.838
.873
.892
.832
.783
.815
.856
Communalities in Factor Analysis:
Communalities indicate the proportion of each variable's
variance that can
be explained by the factors extracted in
the factor analysis.
They are a measure
of the common variance in the data set,
i.e., the variance that is shared
among the variables and
can be accounted for by the underlying factors.nIn
PCA,
initially, the communalities are all set to 1, which means it
is assumed
that all variance is common variance. This is
represented in the "Initial"
column.
Extraction Communalities:nThe "Extraction" communalities show
how
much of the variance in each variable is explained by
the extracted components
after performing PCA These values
range from 0 to 1, where a higher value
indicates that more
of the variance in the variable is explained by the extracted
factors.
Analysis of Your Communalities:nVariables like "Investment in
emerging
technologies" (0.907), "R&D expenditure by
governments and higher
education" (0.892), and "GERD
performed by business enterprise"
(0.873) have high
communalities.
This suggests that a significant portion
of their variances
is explained by the factors extracted in the PCA. These
variables are well represented in the factor solution.nOn the
other hand,
"Mobile apps development" (0.668) and "ICT
skills in the education
system" (0.685) have relatively
lower communalities. While still significant,
this indicates
that these variables have some unique variance that is not
captured by the extracted factors.
Overall, most of your variables have high communalities,
suggesting that the
extracted components do a good job in
explaining a large proportion of the
variance in these
variables.
Explanation of Variance: Communalities show how
much of the
variance in each variable is captured by the entire set of
extracted
factors.
High communalities indicate that the factors explain a large
proportion
of a variable's variance, while lower
communalities suggest that some unique
variance in the
variable is not accounted for by the factors.
Quality of Factor Solution: High communalities across
variables can imply
a good quality factor solution. This
means that the extracted factors are
effectively capturing
the underlying structure or patterns in the data.
Conversely,
if many variables have low communalities, it
might suggest that the factor
solution is not adequately
capturing the data's complexity, or perhaps additional
factors need to be extracted.
Implications for Data Reduction: One of the goals of PCA is
data reduction
– summarizing a large set of variables with a
smaller set of factors
or components.
High communalities suggest that this goal is being met, as
most of the variance in the original variables is being
summarized by the
factors.
In practical terms, this can lead to more efficient and
interpretable models,
especially in contexts where too many
variables can lead to complexity or
overfitting.
Selection of Variables: Communalities can also guide the
selection of variables
for further analysis. Variables with
very low communalities might be considered
for removal since
they do not correlate well with other variables and hence
do
not fit well with the underlying factor structure.
Interpreting Factors: High communalities contribute to the
interpretability
of the factors. If a variable has a high
communality, it means that the factors
explain much of its
variance, and thus, its role in the factor structure is
clearer.
In interpreting the factors, you would focus more on variables
with
higher communalities as they are better represented in
the factor model.
Domain-Specific Insights: In your specific dataset, the
communalities can
provide insights into the nature of the
variables in relation to the underlying
factors.
For example, if variables related to technology investment
have high
communalities, it could indicate that these
investment activities share common
underlying dimensions or
influences.
Limitations: It's important to note that high communalities do
not necessarily
mean that the factor model is the 'true'
model or that it perfectly represents
the data.
Factors are constructs, and while they can offer useful
summaries
and insights, they are one of many possible ways
to understand the data.
Additionally,
communalities do not provide information about
the specific nature of the
underlying factors; for this, you
would need to examine the factor loadings
and interpret the
factors based on these loadings and your understanding of
the variables.
Total Variance Explained
Component
1
2
3
4
5
Initial Eigenvalues
Extraction Sums of Squared Loadings
% of
Cumulative
% of
Cumulative
Total
Total
Variance
%
Variance
%
8.298
51.863
51.863
8.298
51.863
51.863
2.396
14.972
66.835
2.396
14.972
66.835
1.426
8.915
75.751
1.426
8.915
75.751
1.041
6.503
82.254
1.041
6.503
82.254
.613
3.831
86.085
6
.521
3.259
89.344
7
.424
2.650
91.994
8
.357
2.232
94.226
9
.296
1.853
96.079
10
.186
1.160
97.239
11
.135
.841
98.080
12
.108
.676
98.756
13
.091
.571
99.327
14
.057
.355
99.682
15
.044
.277
99.959
16
.007
.041
100.000
Total Variance Explained
Component
1
2
3
4
5
Rotation Sums of Squared Loadings
% of
Cumulative
Total
Variance
%
4.875
30.467
30.467
3.499
21.866
52.333
2.622
16.388
68.720
2.165
13.533
82.254
6
7
8
9
10
11
12
13
14
15
16
Extraction Method: Principal Component Analysis.
Initial Eigenvalues:
Eigenvalues represent the total variance explained by each
principal component.
The first column under "Initial
Eigenvalues" shows the eigenvalue
for each component, which
is a measure of how much of the total variance the
component
captures.
The "% of Variance" column shows the proportion
of the
datasets total variance that each component accounts for.
"Cumulative
%" represents the cumulative variance explained
by the components up
to that point.
Analysis of Components: Component 1: Explains a very high
portion of the variance
(51.863%). This is typical in PCA,
where the first few components explain
the majority of the
variance.
Component 2: Adds another 14.972% to the variance
explained,
bringing the cumulative variance explained to 66.835%.
Component
3: Adds 8.915%, reaching a cumulative 75.751%.
Component 4: Contributes 6.503%,
for a total of 82.254%
variance explained cumulatively. Components 5 to 16:
Each of
these contributes progressively less to the total variance
explained.
Extraction Sums of Squared Loadings: This is similar to the
"Initial
Eigenvalues" but focused on the components that are
retained after the
extraction. In your case, it seems that 4
components have been retained, as
indicated by the non-zero
values in the "Extraction Sums of Squared Loadings"
section.
The variance explained by these four components is the same as
in
the "Initial Eigenvalues" section, as these are the
components extracted.
Rotation Sums of Squared Loadings: After rotation (which is often
done to
make the interpretation of components easier), the
variance explained by each
component can change.
The rotation often spreads the variance more evenly across the
components,
which seems to be the case here: the first
component now explains 30.467%,
and the cumulative variance
explained by the first four components is still
82.254%.
Key Takeaways: The first few components explain a significant
portion of the
total variance. This is a common outcome in
PCA, where the technique aims
to reduce the dimensionality
while retaining as much variance as possible.
The decision on how many components to retain depends on the
cumulative percentage
of variance you aim to explain. Often,
a cut-off is set (like 70%, 80%, etc.),
or a scree plot is
used to identify the point of diminishing returns. The
four
components retained in your analysis explain a total of
82.254% of the
variance, which is quite substantial.
This suggests that these components
capture most of the
underlying structure of your data. It's important to interpret
these components in the context of your variables to
understand what each
component represents.
This involves examining the factor loadings and understanding
how each variable contributes to each component.
Component Matrixa
1
.831
.424
.789
.749
Component
2
3
-.312
-.074
.708
.146
.409
-.074
-.156
.273
4
-.300
.392
-.187
.158
.507
.124
-.232
-.071
-.021
-.141
.100
.436
-.136
.375
.517
-.307
-.012
-.156
-.005
-.196
-.121
-.084
-.291
-.018
.783
.425
-.129
.071
Component
2
3
.035
.353
.921
-.035
.576
.044
.392
.321
4
.286
-.083
-.079
.520
Regulatory quality
AI scientific publications
Robot density
ICT skills in the
education system
Mobile apps
.334
-.532
development
Regulation of emerging
.852
-.206
technologies
Investment in emerging
.926
-.167
technologies
Computer software
.660
.447
spending
GERD financed by
.784
.057
business enterprise
Annual investment in
.253
.643
telecommunication
services
GERD performed by
.887
.248
business enterprise
R&D expenditure by
.893
.235
governments and higher
education
Knowledge intensive
.840
-.324
employment
High-tech exports
.577
-.431
PCT patent applications
.888
-.096
ICT services exports
.141
-.468
Extraction Method: Principal Component Analysis.a
a. 4 components extracted.
Rotated Component Matrixa
Regulatory quality
AI scientific publications
Robot density
ICT skills in the
education system
Mobile apps
development
1
.822
.008
.701
.399
.092
-.045
.199
.786
Regulation of emerging
.706
.194
.522
technologies
Investment in emerging
.774
.270
.381
technologies
Computer software
.162
.864
.236
spending
GERD financed by
.338
.572
.567
business enterprise
Annual investment in
.296
.545
-.662
telecommunication
services
GERD performed by
.745
.538
.147
business enterprise
R&D expenditure by
.776
.515
.130
governments and higher
education
Knowledge intensive
.691
.122
.510
employment
High-tech exports
.188
.117
.836
PCT patent applications
.736
.308
.336
ICT services exports
-.057
-.061
-.084
Extraction Method: Principal Component Analysis.
Rotation Method: Equamax with Kaiser Normalization.
.137
.300
.082
.116
.122
.083
.092
.282
.188
.257
.918
a
a. Rotation converged in 8 iterations.
Understanding the Matrix: Each row represents a variable, and
each column
represents a component (factor). The numbers in
the matrix are the factor
loadings, which indicate the
degree to which each variable correlates with
each factor.
Higher absolute values of the factor loadings (close to 1 or
-1) suggest that the variable strongly correlates with the
factor. Values
closer to 0 suggest a weaker correlation.
Analysis of Factors: Component 1:
Strongly correlated with
"Regulatory quality" (.822), "Robot
density" (.701),
"Investment in emerging technologies" (.774),
and other
similar variables. This factor might represent a general theme
of
technological advancement or regulatory environment.
Component 2: Shows strong
correlation with "AI scientific
publications" (.921) and "Computer
software spending"
(.864). It seems to capture aspects related to scientific
research and development. Component 3: Notably correlated with
"High-tech
exports" (.836) and "GERD financed by business
enterprise"
(.567). This factor might represent
international trade and business investment
in research.
Component 4: Strongly correlates with "ICT services exports"
(.918) and "Mobile apps development" (786), suggesting a focus
on digital services and technology development.
Interpretation: The rotated component matrix helps to clarify
which variables
are most strongly associated with each
component.
This can provide insight
into potential underlying themes or
factors in the data.
For example, if Component
1 is heavily loaded with variables
related to regulatory quality and investment,
it might be
interpreted as a factor representing the 'Regulatory and
Investment
Environment' in technology.
Considerations: The interpretation of the components requires
domain knowledge
and understanding of the context of the
variables.
The labels or themes assigned
to each component are not
generated by the PCA; they are based on the analyst's
interpretation of the pattern of loadings.
Rotation, like Equamax in this case, helps in achieving a
simpler and more
interpretable structure by trying to
maximize the variance among the loadings
for each factor.
Next Steps: The next step in your analysis would be to use
these components
for further analysis, such as clustering,
regression, or other forms of modeling,
depending on your
research objectives.
It's also useful to consider
these components in light of
the original objectives of your analysis to determine
how
they might provide insights or answers to your research
questions.
Identification of Underlying Factors: Each component (factor)
in the matrix
represents an underlying dimension within your
data. The variables that have
high loadings (both positive
and negative) on the same component are considered
to be
measuring the same underlying factor.
Component 1: Technological Advancement & Regulatory
Environment. This
component is strongly associated with
variables like "Regulatory quality,"
"Robot density,"
"Investment in emerging technologies,"
and "R&D expenditure
by governments and higher education."
The
high loadings suggest that this factor represents
elements of technological
advancement and a supportive
regulatory environment.
Component 2: Scientific
Research and Development. With high
loadings on "AI scientific publications"
and "Computer
software spending," this factor seems to capture the
research and development aspect, particularly in the fields of
AI and software.
It indicates a focus on scientific innovation and
technological research.
Component 3: International Trade and
Business Investment. This component is
characterized by high
loadings on "High-tech exports" and "GERD
financed by
business enterprise," indicating a focus on international
technology trade and business investment in research and
development.
Component
4: Digital Services and Technology Development.
Strongly correlated with "ICT
services exports" and "Mobile
apps development," this factor
likely represents the digital
services sector and the development of mobile
technologies.
Interrelationships among Variables: The matrix helps
understand
how different variables are interrelated. For
instance, if two variables load
highly on the same factor,
it suggests they share a common underlying theme
or
construct.
Simplification for Further Analysis: The components provide
a simplified structure of the data. Instead of dealing with
numerous individual
variables, you can now work with a
smaller number of factors that represent
combinations of
these variables.
Guidance for Strategic Decisions: For policy
makers,
business strategists, or researchers, understanding these
factors
can guide strategic decisions.
For example, if a factor representing technological
advancement is important, focusing on variables that load
highly on this factor
could be beneficial.
Limitations - Interpretation Subjectivity: The interpretation
of factors is
subjective and depends on the context and
domain knowledge. Different analysts
might interpret the
same factor loadings in different ways.
Rotation Effectiveness:
The rotation method (Equamax in your
case) has successfully simplified the
factor structure,
making it easier to interpret the components.
This is evident
from the distinct groupings of variables on
different components.
Component Transformation Matrix
1
2
3
4
Component
1
.726
.494
.397
.268
2
-.021
.704
-.494
-.509
3
-.155
.212
-.526
.809
4
-.670
.464
.568
.119
Extraction Method: Principal Component Analysis.
Rotation Method: Equamax with Kaiser Normalization.
Means
Case Processing Summary
REGR factor score
analysis 1 * Group
REGR factor score
analysis 1 * Group
REGR factor score
analysis 1 * Group
REGR factor score
analysis 1 * Group
1 for
Included
N
Percent
36 100.0%
Cases
Excluded
N
Percent
0
0.0%
Total
N
Percent
36 100.0%
2 for
36
100.0%
0
0.0%
36
100.0%
3 for
36
100.0%
0
0.0%
36
100.0%
4 for
36
100.0%
0
0.0%
36
100.0%
Report
Group
Candidate Mean
N
Std.
Deviation
Variance
Kurtosis
Skewness
Emerging Mean
N
Std.
Deviation
Variance
Kurtosis
Skewness
Innovator Mean
s
N
Std.
Deviation
REGR factor REGR factor REGR factor REGR factor
score 1 for
score 2 for
score 3 for
score 4 for
analysis 1
analysis 1
analysis 1
analysis 1
-.7827035
-.4453512
-.8253943
-.3131225
9
9
9
9
.62597466
1.00725859
.55912889
1.03855649
.392
.320
-.664
-.2004172
16
.69061839
1.015
.333
1.338
-.0255513
16
.97932620
.313
-.851
-.204
.2732813
16
1.22564017
1.079
-.144
.395
-.2048532
16
.62801664
.477
.314
.161
.9319098
11
.94876914
.959
-.810
.503
.4015437
11
.94488876
1.502
13.429
3.549
.2778226
11
.42316044
.394
-.205
.454
.5541594
11
1.24592383
Total
Variance
Kurtosis
Skewness
Mean
N
Std.
Deviation
Variance
Kurtosis
Skewness
.900
.129
-1.110
.0000000
36
1.00000000
.893
1.010
-.649
.0000000
36
1.00000000
.179
-1.296
.485
.0000000
36
1.00000000
1.552
-.362
.912
.0000000
36
1.00000000
1.000
-.680
.299
1.000
-1.101
.271
1.000
14.034
2.878
1.000
1.417
1.034
ANOVA Tablea
REGR factor score 1 for Between
analysis 1 * Group
Groups
Within Groups
Total
REGR factor score 2 for Between
analysis 1 * Group
Groups
Within Groups
Total
REGR factor score 3 for Between
analysis 1 * Group
Groups
Within Groups
Total
REGR factor score 4 for Between
analysis 1 * Group
Groups
Within Groups
Total
(Combined)
Sum of
Squares
15.709
(Combined)
19.291
35.000
3.569
33
35
2
(Combined)
31.431
35.000
8.175
33
35
2
(Combined)
26.825
35.000
4.932
33
35
2
30.068
35.000
33
35
df
2
ANOVA Tablea
REGR factor score 1 for Between
analysis 1 * Group
Groups
Within Groups
(Combined)
Mean
Square
7.855
F
13.437
Sig.
<.001
1.874
.170
.585
Total
REGR factor score 2 for Between
analysis 1 * Group
Groups
Within Groups
(Combined)
1.785
.952
Total
REGR factor score 3 for Between
analysis 1 * Group
Groups
Within Groups
(Combined)
4.088
5.029
.012
2.706
.082
.813
Total
REGR factor score 4 for Between
analysis 1 * Group
Groups
Within Groups
(Combined)
2.466
.911
Total
a. The grouping variable Group is a string, so the test for linearity cannot be computed.
Measures of Association
REGR factor score
analysis 1 * Group
REGR factor score
analysis 1 * Group
REGR factor score
analysis 1 * Group
REGR factor score
analysis 1 * Group
1 for
Eta
.670
Eta
Squared
.449
2 for
.319
.102
3 for
.483
.234
4 for
.375
.141
Univariate Analysis of Variance
Between-Subjects
Factors
N
Group Candidate
Emerging
Innovator
s
9
16
11
Levene's Test of Equality of Error Variancesa,b
REGR factor score 1 for Based on Mean
analysis 1
Based on Median
Based on Median and
with adjusted df
Based on trimmed mean
Levene
Statistic
1.150
.313
.313
df1
df2
2
2
2
33
33
24.182
.927
2
33
F
13.437
Sig.
<.001
.017
13.437
.897
<.001
Levene's Test of Equality of Error Variancesa,b
REGR factor score 1 for Based on Mean
analysis 1
Based on Median
Based on Median and
with adjusted df
Based on trimmed mean
Sig.
.329
.734
.734
.406
Tests the null hypothesis that the error variance of the
dependent variable is equal across groups.a,b
a. Dependent variable: REGR factor score 1 for analysis 1
b. Design: Intercept + Group
Tests of Between-Subjects Effects
Dependent Variable: REGR factor score 1 for analysis 1
Type III Sum
Mean
of Squares
df
Square
Source
a
Corrected
15.709
2
7.855
Model
Intercept
.010
1
.010
Group
15.709
2
7.855
Error
19.291
33
.585
Total
35.000
36
Corrected Total
35.000
35
a. R Squared = .449 (Adjusted R Squared = .415)
Post Hoc Tests
Group
Multiple Comparisons
Dependent Variable: REGR factor score 1 for analysis 1
Tukey HSD
95% Confidence Interval
Mean
Difference (ILower
Upper
J)
Std. Error
Sig.
Bound
Bound
(I) Group (J) Group
Candidate Emerging
-.5822863 .31857058
.176 -1.3639930
.1994205
*
Innovator
-1.7146133 .34364838
<.001 -2.5578558
-.8713708
s
Emerging Candidate
.5822863 .31857058
.176
-.1994205 1.3639930
*
Innovator
-1.1323270 .29946251
.002 -1.8671465
-.3975076
s
Innovator Candidate
1.7146133* .34364838
<.001
.8713708 2.5578558
*
s
Emerging
1.1323270 .29946251
.002
.3975076 1.8671465
Based on observed means.
The error term is Mean Square(Error) = .585.
*. The mean difference is significant at the .05 level.
Homogeneous Subsets
REGR factor score 1 for analysis
1
Tukey HSDa,b,c
Subset
Group
Candidate
Emerging
N
1
9 -.7827035
16 -.2004172
2
Innovator
s
Sig.
11
.9319098
.181
1.000
Means for groups in homogeneous subsets
are displayed.
Based on observed means.
The error term is Mean Square(Error) =
.585.
a. Uses Harmonic Mean Sample Size =
11.341.
b. The group sizes are unequal. The
harmonic mean of the group sizes is used.
Type I error levels are not guaranteed.
c. Alpha = .05.
Univariate Analysis of Variance
Between-Subjects
Factors
N
Group Candidate
Emerging
Innovator
s
9
16
11
Levene's Test of Equality of Error Variancesa,b
REGR factor score 2 for Based on Mean
analysis 1
Based on Median
Based on Median and
with adjusted df
Based on trimmed mean
Levene
Statistic
.268
.322
.322
.266
Levene's Test of Equality of Error Variancesa,b
Sig.
df1
df2
2
2
2
33
33
29.425
2
33
REGR factor score 2 for Based on Mean
analysis 1
Based on Median
Based on Median and
with adjusted df
Based on trimmed mean
.767
.727
.727
.768
Tests the null hypothesis that the error variance of the
dependent variable is equal across groups.a,b
a. Dependent variable: REGR factor score 2 for analysis 1
b. Design: Intercept + Group
Tests of Between-Subjects Effects
Dependent Variable: REGR factor score 2 for analysis 1
Type III Sum
Mean
df
of Squares
Square
Source
a
Corrected
3.569
2
1.785
Model
Intercept
.018
1
.018
Group
3.569
2
1.785
Error
31.431
33
.952
Total
35.000
36
Corrected Total
35.000
35
a. R Squared = .102 (Adjusted R Squared = .048)
Post Hoc Tests
Group
Multiple Comparisons
F
1.874
Sig.
.170
.019
1.874
.891
.170
Dependent Variable: REGR factor score 2 for analysis 1
Tukey HSD
95% Confidence Interval
Mean
Difference (ILower
Upper
Std. Error
Sig.
J)
Bound
Bound
(I) Group (J) Group
Candidate Emerging
-.4197999 .40664012
.562 -1.4176112
.5780114
Innovator
-.8468949 .43865074
.146 -1.9232537
.2294638
s
Emerging Candidate
.4197999 .40664012
.562
-.5780114 1.4176112
Innovator
-.4270950 .38224958
.510 -1.3650569
.5108669
s
Innovator Candidate
.8468949 .43865074
.146
-.2294638 1.9232537
s
Emerging
.4270950 .38224958
.510
-.5108669 1.3650569
Based on observed means.
The error term is Mean Square(Error) = .952.
Homogeneous Subsets
REGR factor score 2 for
analysis 1
Tukey HSDa,b,c
Group
Candidate
Emerging
Innovator
s
Sig.
Subset
N
1
9 -.4453512
16 -.0255513
11 .4015437
.113
Means for groups in
homogeneous subsets are
displayed.
Based on observed means.
The error term is Mean
Square(Error) = .952.
a. Uses Harmonic Mean Sample
Size = 11.341.
b. The group sizes are unequal.
The harmonic mean of the group
sizes is used. Type I error levels
are not guaranteed.
c. Alpha = .05.
Univariate Analysis of Variance
Between-Subjects
Factors
N
Group Candidate
Emerging
Innovator
s
9
16
11
Levene's Test of Equality of Error Variancesa,b
REGR factor score 3 for Based on Mean
analysis 1
Based on Median
Based on Median and
with adjusted df
Based on trimmed mean
Levene
Statistic
.403
.209
.209
.199
Levene's Test of Equality of Error Variancesa,b
REGR factor score 3 for Based on Mean
analysis 1
Based on Median
Based on Median and
with adjusted df
Based on trimmed mean
Tests the null hypothesis that the error variance of the
dependent variable is equal across groups.a,b
Sig.
.671
.813
.814
.820
df1
df2
2
2
2
33
33
17.785
2
33
a. Dependent variable: REGR factor score 3 for analysis 1
b. Design: Intercept + Group
Tests of Between-Subjects Effects
Dependent Variable: REGR factor score 3 for analysis 1
Type III Sum
Mean
of Squares
df
Square
Source
a
Corrected
8.175
2
4.088
Model
Intercept
.284
1
.284
Group
8.175
2
4.088
Error
26.825
33
.813
Total
35.000
36
Corrected Total
35.000
35
F
5.029
Sig.
.012
.350
5.029
.558
.012
a. R Squared = .234 (Adjusted R Squared = .187)
Post Hoc Tests
Group
Multiple Comparisons
Dependent Variable: REGR factor score 3 for analysis 1
Tukey HSD
95% Confidence Interval
Mean
Difference (ILower
Upper
J)
Std. Error
Sig.
Bound
Bound
(I) Group (J) Group
Candidate Emerging
-1.0986756* .37566268
.017 -2.0204746
-.1768766
*
Innovator
-1.1032170 .40523477
.027 -2.0975798
-.1088541
s
Emerging Candidate
1.0986756* .37566268
.017
.1768766 2.0204746
Innovator
-.0045414 .35313019
1.000
-.8710503
.8619676
s
Innovator Candidate
1.1032170* .40523477
.027
.1088541 2.0975798
s
Emerging
.0045414 .35313019
Based on observed means.
The error term is Mean Square(Error) = .813.
*. The mean difference is significant at the .05 level.
Homogeneous Subsets
REGR factor score 3 for analysis
1
Tukey HSDa,b,c
Subset
Group
Candidate
N
1
9 -.8253943
2
Emerging
16
.2732813
Innovator
s
Sig.
11
.2778226
1.000
1.000
Means for groups in homogeneous subsets
are displayed.
Based on observed means.
The error term is Mean Square(Error) =
.813.
a. Uses Harmonic Mean Sample Size =
11.341.
b. The group sizes are unequal. The
harmonic mean of the group sizes is used.
Type I error levels are not guaranteed.
c. Alpha = .05.
Univariate Analysis of Variance
Between-Subjects
1.000
-.8619676
.8710503
Factors
N
Group Candidate
Emerging
Innovator
s
9
16
11
Levene's Test of Equality of Error Variancesa,b
REGR factor score 4 for Based on Mean
analysis 1
Based on Median
Based on Median and
with adjusted df
Based on trimmed mean
Levene
Statistic
2.861
1.082
1.082
2.524
df1
df2
2
2
2
33
33
21.672
2
33
Levene's Test of Equality of Error Variancesa,b
REGR factor score 4 for Based on Mean
analysis 1
Based on Median
Based on Median and
with adjusted df
Based on trimmed mean
Sig.
.071
.351
.356
.095
Tests the null hypothesis that the error variance of the
dependent variable is equal across groups.a,b
a. Dependent variable: REGR factor score 4 for analysis 1
b. Design: Intercept + Group
Tests of Between-Subjects Effects
Dependent Variable: REGR factor score 4 for analysis 1
Type III Sum
Mean
df
of Squares
Square
Source
a
Corrected
4.932
2
2.466
Model
Intercept
.005
1
.005
F
2.706
Sig.
.082
.005
.942
Group
Error
4.932
30.068
2
33
Total
35.000
36
Corrected Total
35.000
35
2.466
.911
2.706
.082
a. R Squared = .141 (Adjusted R Squared = .089)
Post Hoc Tests
Group
Multiple Comparisons
Dependent Variable: REGR factor score 4 for analysis 1
Tukey HSD
95% Confidence Interval
Mean
Difference (ILower
Upper
J)
Std. Error
Sig.
Bound
Bound
(I) Group (J) Group
Candidate Emerging
-.1082693 .39772694
.960 -1.0842095
.8676709
Innovator
-.8672819 .42903592
.123 -1.9200479
.1854841
s
Emerging Candidate
.1082693 .39772694
.960
-.8676709 1.0842095
Innovator
-.7590126 .37387102
.121 -1.6764153
.1583900
s
Innovator Candidate
.8672819 .42903592
.123
-.1854841 1.9200479
s
Emerging
.7590126 .37387102
.121
-.1583900 1.6764153
Based on observed means.
The error term is Mean Square(Error) = .911.
Homogeneous Subsets
REGR factor score 4 for
analysis 1
Tukey HSDa,b,c
Group
Candidate
Emerging
Innovator
s
Sig.
Subset
N
1
9 -.3131225
16 -.2048532
11 .5541594
.093
Means for groups in
homogeneous subsets are
displayed.
Based on observed means.
The error term is Mean
Square(Error) = .911.
a. Uses Harmonic Mean Sample
Size = 11.341.
b. The group sizes are unequal.
The harmonic mean of the group
sizes is used. Type I error levels
are not guaranteed.
c. Alpha = .05.
Download