SPSS has highlighted two cases in my survey questions where a horizontal line appears in boxplot. The screenshot of one such case is attached. For those that do not fall within the line, it is suggested that they could be outliers. I had checked the questions for potential errors and there were none. What should I do?
It seems that your data is not continuous (or heavily tied). You should better treat it as discrete ordinal data.
A boxplot includes the central 50% of the values in the box ("interquartile range"!). Obviousely, 15 of your 30 values are "4". There is no finite width of the interquatile range - its width is zero. So the plot is correct, but the kind of data is inappropriate for this kind of plot. You may use a histogram instead (i.e. a barplot showing the [relative] frequencies).
Yes, the histogram looks acceptable to a bell-shape. This question is part of a survey that used continues ordinal data (5 point likert scale) for all its questions. Could I presume that the graphics in the botplot is not something, I should be concerned with or could it suggest I should pursue additional statistics testing or validation?
Values from a Likert scale are ordinal values. For exploration, plotting and inference you should use appropriate methods (for ordinal data). You will also find a lot of literature especially dealing with Likertd-scaled data.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS