Factor analysis is used to reduce and summarize the data being analyzed , eliminating dependencies or independence from the data. It is also used to "identify factors that explain the correlations between the variables" , being that in this type of analysis all the variables are equal, that is to say independent, which means that no variable is more or less than another .

Having the required information, it is necessary to calculate a set of latent dimensions called factors, with which these relationships are explained. Therefore, factor analysis is a data reduction technique with which the hypotheses put forward can be tested.

Factorial Analysis can be of two types. The exploratory type, in which the number of factors is not known from the beginning and it is in the empirical application where this number is determined. In the confirmatory analysis, the factors are known from the outset, using the hypotheses for corroboration.

Hello!

The use simple sum of scores is part of "Classical Test Theory". More modern approach is known as "Item Response Theory", which is also called "Logistic factor analysis".

Article Overview of Classical Test Theory and Item Response Theory f...

You can explore your data by using "Multiple Correspondence Analysis" (example http://factominer.free.fr/factomethods/multiple-correspondence-analysis.html ).

http://www.personality-project.org/r/psych/ - R package with examples.

Dear Janila,

In order to analyze Likert scales, you may need to follow these simple guide:

1. Was the scale you used being validated? If yes, get the original scoring methods from the owner. Likert scales can be tricky here, the items in the scale may be negatively worded or positively worded based on the logical characteristic of the hypothesized main outcome measure. If it's the former, you need to reverse the scores on that particular item first.

2. Compute the total scores. It's just as simple as tabulating each individual item score to have a new variable of the total score!

3. Next determine your tests by appreciating the nature of your variable. Since it's total scores, your variable is continuous in nature. Thus inferential statistics may use either correlational analysis, t-tests or ANOVA. However, if you wish to dichotomize your score of the dependent variable, then you need to use chi-square test or simple logistic regressions depending on the nature of your independent variables.

Janila Joven. At the most basic level in answer to your question, I think you should not round off your composite / total score. It actually lies mid-way between Strongly disagree and Disagree, so I think it represents quite well where the particular respondent you give the example of lies.

Some of the other answers provided above do not really address your question, but they could well be worth considering in terms of issues relating to the reliability and validity of the scale you are using.

Strictly speaking, you cannot assume that the distance between, say, strongly agree and agree is the same as the distance between, say, agree and neutral. For example, better values might be -3, -1, 0, 1, and 3, but there is no way to know from the information you presented. Thus, calculating averages is not appropriate. There are ways of creating scales, but you would need more information than you present here. Alternatively, you can treat this as a categorical (in this case, ordinal) variable and use techniques such as chi-square (a variety of tools are available), or convert it into a dichotomous variable (for which averages may be calculated), such as agree/strongly agree vs. all other responses, or disagree/strongly disagree vs. all other responses..

Article Scale quality: alpha is an inadequate estimate and factor-an...