Day 8 : Team G

Factor analysis: It is a statistical method to describe the variability among observed and correlated variables. It helps reduce the number of variables by finding the commonality between the factors/attributes and thus determine the common underlying theme. It is one of the techniques for data reduction. 

It is basically of 2 types:

·         Exploratory factor analysis

·         Confirmatory factor analysis

Primarily used to:

1.       Reduce the number of variables

2.       Find the common underlying theme

Factor analysis helps in the identification of the underlying factors by clustering variables and creating new variables. In addition to it, it helps in the screening of the variables and gives us fewer variables to represent the entire set.

Note: String variables cannot be used for factor analysis

To perform factor analysis in SPSS,

Analyse> Data reduction> Factor analysis

Make sure that the following parameters are selected:

·         Initial solution

·         Scree plot

·         Eigen value>1

·         Verimax

Advantages	Disadvantages
Both subjective and objective attributes could be used	Large number of product attributes required
Can identify dimensions which direct analysis cannot	Similarity in observed variables creates difficulty in exact representation
Easy and inexpensive

Scree plot:

It gives the plot of the Eigen values of the correlation matrix, in the descending order of magnitudes. It helps visualise the relative importance of the factors.

Normalization:

We need to normalize the variables so that we have a common platform for comparison/ same measurement level. This is done by computing the Z-score.

It is computed as:

Z-score = variation/std deviation

From the 1^st table (Communalities), the “initial” value gives us the variance and the extraction value in each variable. More the extraction value, better it is. If a minimum value of 0.5 is not present, drop that particular variable.

Extraction:

The process by which the commonality between factors/variables is shown. Higher value of extraction implies higher overlap/ commonality. Consider overlapping the Z-score with the original. The area of overlap between both indicates the extraction value.

Author:

Anand Chandran