Adjusted R-square
Analysing qualitative data.
Analysing quantitative data
ANCOVA (Analyses of covariance)
ANOVA (Analyses of variance)
Applied research
Average
Bias
[...]
Drop outs
A drop out is a deliberately deleted object or respondent by the researcher. To delete objects or respondents is always disputable.
Is it reasonable to have drop outs?
Objects or respondents can be removed from the database. It is always questionable to remove a respondent or an object. If a respondent does not understand what the research is all about and the answers to the questions don’t make sense, then it is clear the answers of this respondent can be removed. However, this is easily said but usually this is not always clear. Let us discuss a number of situations.
Outlier
An outlier is a respondent or an object with extreme values. When a mean is calculated, it will be higher (or lower) than when the mean is calculated without this data. These extreme values however might be true, and removing these data is violating the reality in favour of the statistics. On the other hand, if these data aren’t deleted a total different view on the subject of your research might be the result. What can be done against this?
There are some options. First of all, if there are enough respondents, the impact of an outlier will not be too large, it won´t have much impact. Secondly, try using non-parametric statistics, because these types of analyses are not affected by extreme values. Thirdly, do the analyses once with and once without outliers. If totally different conclusions are drawn, outliers have a big impact.
Drop outs are sometimes removed by the provider
In online surveys a provider is involved to gather the data. Providers do not always provide all data. A while ago it was very common that only the data of the respondents who had fully completed the questionnaire were supplied. It is okay to drop the respondents who clicked on the button to start the questionnaire but did not answer any question. However, if respondents answered all the questions except the final two, should they be deleted? There is no reason to doubt whether the previous answers are true. According to me, this should not be a reason for deleting all data of this respondent.
Illegal drop outs
Sometimes respondents are deleted because they do not fit in nicely, that is, these respondents are not supporting the theory of the researcher. For instance, rats raised in the laboratory were given a toxic infusion and after a four weeks, they were killed and the amount and severity of the tumours were counted. The rats who passed away in the first four weeks, were deleted from the data. Consequently, this gives an incomplete view on the toxicity of the infusion.
Likewise, data of the severely polluted areas can be omitted or even not gathered. Not collecting the data is not really a dropout (it is not afterwards but beforehand), but it is a severe form of not being able to draw the right conclusions from the data.
Related topics to drop outs:
Do professional research. Read these manuals:
