Factor analysis is a type of statistical analysis that investigates different correlations and patterns that may occur between measurements. There are two types of factor analysis; exploratory and confirmatory. These two versions may be used individually or combined. There are many different types of statistical calculations that are used within this analysis.
A common first step used in factor analysis includes collecting the measurements in the experiment. The correlation mathematics are used to determine existing correlations. The researcher will determine if all the factors calculated from the analysis will be included. Some experiments will require certain factors to be incorporated into the statistics and others to be excluded.
One method that is used to extract the possible factors is maximum likelihood. This calculation is so complicated that statistical computer programs are used, as a researcher typically cannot perform the calculation by hand. The factors within the analysis may also be combined in a number of ways. The analysis will require the order of the factors to be rotated or combed in a way that explains the greats variance or spread of data.
Once the final factors and scores are calculated, the data can be interpreted. Factors that have the highest scores will have the most influence on the measurements. These scores may also be used for further statistical analysis. Unlike other types of statistical analysis, this analysis can result in an unlimited amount of important factors, rather than restricting the factors to a small group.
Exploratory factor analysis is used to understand which things in nature can influence certain measurements. How strongly these factors influence the measurements is also of interest in the exploratory version. These are not preset before the measurements are taken. With confirmatory factor analysis, there are specific factors that are being investigated prior to the calculations.
Both types of factor analysis can be used within one experiment. The exploratory version may be used to create a theory, while the confirmatory version is used to prove that theory. If the confirmatory analysis is not favorable, then the researcher may need to change how the exploratory analysis is calculated.
The number of measurements required for these calculations are important. Most calculations require at least ten measurements if not more. Usually confirmatory analysis will need many more measurement than exploratory. At times, at least 200 measurements are needed for a successful analysis. As a general rule, using more measurements typically results in more reliable data, though the necessary number will depend on the experiment.
