Editing 3.4:Hypothesis Testing

{{av}}

You can test various hypotheses using content analysis. However, it is important that you formulate these hypotheses clearly and concretely prior to performing the content analysis. Using content analysis, you can extrapolate your findings (i.e. generalization them to a broader context, which you have not actually measured). Examples of such extrapolations are trends, patterns and differences. From these trends, patterns and differences you observe in the collection of documents that you analyzed (your sample), you can draw conclusions about similar documents that you have not actually analyzed (the population) and thus test your hypotheses. Each of these extrapolation methods is discussed below. Please note that although these three extrapolation methods are discussed seperately, things tend to get ''really'' interesting when you start combining different extrapolation methods! 

==Trends==

A trend is an association between time and a measurement. A trend hypothesis argues that this measurement depends on time. AmCAT provides you with various possibilities to analyse trends. Using AmCAT, you can visualize a trend graphically any draw preliminary conclusions about the trend. You could, for example, test whether a trend is linear or cyclic. To display a possible trend in a graph you can use the graph function in the [[3.4:Graph/Table|Graph/Table function]]. AmCAT also provides you with the possibility to produce data with which you can perform a formal test of the trend hypothesis. To test whether a certain trend exists (i.e. whether a measurement depends on time), you export the AmCAT output by clicking 'Export Aggregation' and analyse the data in a statistical program.

==Patterns==

A pattern is a text element or a combination of text elements with a high occurrence. An association pattern indicates that certain concepts often co-occur. You can measure an association pattern by looking for co-occurrence of search terms within a certain unit of analysis. Using AmCAT, you can visualize such association patterns, but you can also use AmCAT to perform formal statistical tests with regard to patterns. To test patterns you can use the [[3.4:Summary|Summary function]]. Using the 'ClusterMap' option, you can visualize association patterns. The 'Associations' option offers various opportunities for quantifying the strength of associations between concepts. You can display your results graphically or in a table. To test whether certain concepts co-occur more often than expected, you export the AmCAT output by clicking 'Export Aggregation' and analyse the data in a statistical program.

==Differences==

Lastly, you can extrapolate your findings based on the differences observed between certain units. You could, for example, compare articles from different media/sources, debates within the field of politics versus media, attention for a concept in period A verus B, et cetera. Using AmCAT, you can look for differences between certain units using the [[3.4:Graph/Table|Graph/Table function]] or the 'associations' option in the [[3.4:Summary|Summary function]]. To test whether the differences between units are significant, you export the AmCAT output by clicking 'Export Aggregation' and analyse the data in a statistical program.