Data interpretation

The questions as to when values should be interpreted as low or high and which conclusions should be drawn from the collected data depend on the circumstances and questions of the individual project. A generalisation is therefore not possible. (This applies especially to the data that is gathered using content-analytical methods.) Nevertheless, we would like to give some examples.

Scale (mean) values from questionnaire data can be, for example, interpreted according to the meaning of the scale description. For example “Students assess xy on average as 'good'“, etc. Interpretation of results can also occur by an approach of comparison:

• among variables
  
• among student groups
• among different points in time
  
• among courses / classes
• among software tools

Example:

The questionnaire on learning-related expectations of personal effectiveness is again used to illustrate the point. We assume that this questionnaire has been filled out twice by the same sample: first before and then after a learning skill training. The question here is whether the expectation of personal effectiveness in learning has improved after the training.

We enter the data of the second questionnaire survey as new variables (column) into the data sheet of the first survey and calculate for each participant and for each time of measurement a mean value over the questions of expectations of self-efficacy. A t-test is conducted to check whether the mean values of the after-measurement differ from those of the before-measurement.

Sample

The results show that the mean values are higher after the learning skill training (cf. t-test example, pdf, 20 kB, in German) and the difference between the two scale mean values (sk_1 = 3.15 and sk_2 = 4.01) is significant (p < 0.01). This means that the measurement after the training has resulted in an improvement which is highly unlikely to be coincidental .

In the interpretation of the values, the size and the composition of the sample should also be considered. A small number of participants can bias the results. The representativeness of the sample should also be considered. This means that the sample should reflect the composition of the population. Representativeness can be achieved, for example, by a random selection of the survey participants.

Interpretation of correlations / associations

Correlations – irrespective of their levellevel - do not say anything about causalities. Whether a leads to b or vice-versa cannot be explained by correlational methods.