Assignment #1: FACTOR ANALYSIS ON satisf.

sav
KMO & Bartlett's Test

1.

After applying factor analysis on the given data we found that the determinant is 0.058 i.e. determinant > 0.0001. Thus by considering KMO & Bartlett's Test, we found that the test is significant and fitted as P-Value < 0.05. Approximately 78.8% of the actual data of 17 variables will be represented by 7 constructed factors.

KMO and Bartlett's Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy. Bartlett's Test of Sphericity Approx. Chi-Square df Sig. .788

1489.641 136 .000

2.

DETERMINANT
Correlation Matrixa a. Determinant = .058

3.

ROTATED COMPONENT MATRIXA

Rotated sums of squares showed that around 63% of the actual variance is explained by these constructed factors. By observing the Rotated Component Matrix we can identify that the belonging of the variables to any of the factors as follows:

Rotated Component Matrixa

satis Gender Age category Shopping frequency Regular reason Primary department Made purchase Method of payment Follow-up via Distance from home Store Contact with employee Price satisfaction Variety satisfaction Organization satisfaction Service satisfaction Item quality satisfaction Overall satisfaction Dist Component empcnt Shop Demo .730 .677 .665 .388 .769 .823 -.686 .820 .809 .838 .833 .834 .684 .799 .692 .796 Conct Age .865

% of variance Reliability

19.64% .860

8.33% .554

7.51% .262

7.17% .375

6.88% 0.11298

6.55% -.154

6.51% Nil

Based on the information gathered by Rotated Component Matrixa, we have to compute the factors as variables for further decision making analysis. The most representative and reliable factor is satis with 86% reliability, while least reliable factor is demo with only 11% reliability. The explanation and reliability of other factors is also mentioned in the above table.

Assignment #2: Factor Analysis and Regression on world95.sav

FACTOR ANALYSIS
1.

KMO & Bartlett's Test

After applying factor analysis on the given data we found that the KMO & Bartlett's Test is significant and test is fitted as P-Value < 0.05 and determinant < 0.0001. Approximately 77.5% of the actual data of 24 variables will be represented by 5 constructed factors.

KMO and Bartlett's Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy. Bartlett's Test of Sphericity Approx. Chi-Square df Sig. .775

1935.139 276 .000

2.

DETERMINANT
Correlation Matrixa

a. Determinant = 8.07E-018

3.

ROTATED SUMS OF SQUARES

Rotated sums of squares showed that around 82% of the actual variance is explained by these constructed factors. By observing the Rotated Component Matrix we can identify that the belonging of the variables to any of the factors as follows:

Rotated Component Matrixa

LIT Population in thousands Number of people / sq. kilometer People living in cities (%) Average female life expectancy Average male life expectancy People who read (%) Population increase (% per year)) Infant mortality (deaths per 1000 live births) Gross domestic product / capita Region or economic group Daily calorie intake Aids cases Birth rate per 1000 people Death rate per 1000 people Number of aids cases / 100000 people Log (base 10) of GDP_CAP Log (base 10) of AIDS_RT Birth to death ratio Fertility: average number of kids Log (base 10) of Population cropgrow Males who read (%) Females who read (%) Predominant climate LifeStats Component Disease ECO .422 .742 .933 .899 .950 .773 -.941 .806 -.686 .659 .649 -.853 -.746 .833 .719 .820 .829 -.851 .878 -.582 .899 .946 -.709 POP .802

% of variance Reliability

39.12 0.127

13.22

10.18 0.001

10.13 0.059

8.98

The most representative and reliable factor is LIT with 12.7% reliability, while least reliable factor is ECO with only 5.9% reliability. The explanation and reliability of other factors is also mentioned in the above table.

REGRESSION ANALYSIS
Develop a model for ECO of the country based on POP, LIFESTATS, DISEASE and LIT. Independent: Eco Dependant: POP, LIFESTATS, DISEASE and LIT (4 variables)

1.

MODEL SUMMARY

The new variables may be computed by these factors and then regression will be applied by taking ECO as dependent variable as follows. The below summary is showing that we have a strong relation (79.5%) of our dependent variable with all the independent variables and we can (63.3%) accurately predict/explain it with the help of independent variables and 1045 is the standard error of the estimate. No auto correlation is found as Durbin-Watson 2. Model Summaryb Model 1 R .795a R Square .633 Adjusted R Square .605 Std. Error of the Estimate 1045.37507 DurbinWatson 1.824

a. Predictors: (Constant), Disease, lifestats, lit, pop b. Dependent Variable: eco

2.

ANOVA Table

The ANOVA table is showing that our regression model is significant since Sig. Value < 0.05 and we can use it for further analysis. ANOVAb Model Regression Residual Total Sum of Squares 1.016E8 5.901E7 1.607E8 df 4 54 58 Mean Square 2.541E7 1092809.046 F 23.254 Sig. .000a

a. Predictors: (Constant), Disease, lifestats, lit, pop b. Dependent Variable: eco

3.

COEFFICIENTS

The table is showing that the impact of pop and lifestats is found insignificant on the eco. No Multi Collinearity is found since VIF < 10 and Tolerance > 0.1. Coefficientsa Model Unstandardized Standardized Coefficients Coefficients B Std. Error Beta -3618.107 660.429 -.002 .001 -.099 18.910 2.680 .600 -23.596 33.888 -.060 .036 .008 .392 t Sig. Collinearity Statistics Tolerance -5.478 -1.150 7.056 -.696 4.568 .000 .255 .000 .489 .000 .925 .941 .929 .923 VIF 1.081 1.063 1.076 1.084

(Constant) pop lit lifestats Disease

a. Dependent Variable: eco

Analyzing impacts of variables we found that: 1 unit increase in POP decreases 0.002 units in ECO 1 unit increase in LIT increases 18.910 units in ECO 1 unit increase in LIFESTATS decreases 23.596 units in ECO 1 unit increase in DISEASE increases 0.036 units in ECO

The regression equation will be expressed as;

ECO = -3618.107 - 0.002 (POP) + 18.910 (LIT) - 23.596 (LIFESTATS) + 0. 036 (DISEASE)

4.

SCATTER PLOT

By considering the scatter diagram between the dependant variable (eco) and square residuals generated by regression analysis, we didnt find any pattern representing the hetroscedasity. Thus we conclude that the data is free from hetroscedasity.

CONCLUSION
After applying factor analysis based on variables we performed regression analysis keeping eco as dependant variable and the remaining four variables as independent variables. We found that this model can produce 79.5% accuracy to predict the economy of a country, free from auto correlation, multi collinearity and hetroscadasity.

Quantative Techniques in Analysis

ASSIGNMENTS

Submitted to

Syed Ali Raza

Esha Nadeem 6429 MBA (E) Spring 2013