I am writing a discussion for my study and I am trying to highlight any limitations. I have already added demand characteristics due to questionnaires, and having a small sample size.
I did the relationship between age and Emotional intelligence, and the participants age range was 19-45. I did find a positive correlation between the two. However I have noticed that there is an uneven distribution of age as 86% percent of the participants were between ages 19-21. I understand that this could be a limitation however I do not know how to explain it. How can having an uneven distribution of data affect my study and results? How can I explain this point?
You are right to be suspicious! With a small number of people outside the range 19-21, those high values can have a considerable leverage, tilting the regression line in their direction.
Have a look at a scatterplot with a regression line and a lowess line. This is my usual first stop when looking at data like this.
As a matter of interest, why are you testing the relationship in such an odd sample?
I would recommend a proper controlled random sample in a design-based paradigm. But if this is all you got, then you have a big cloud of points in the lower range & a small number of points in the upper range. The small number of points towards the end of the age range will tend to define the slope of your line. Ronan is correct you need to look at a scatter plot. Another thought is instead of a regression define age groupings & try unbalanced ANOVA.
Whatever results you get they are liable to be biased when extended to any generic population. I am not sure any scientific valid results can be obtained from such a sample.
The unfortunate thing is that this population might be students (it's a familiar age distribution to anyone in education!), in which case the age distribution is correct, and attempts to oversample older students will result in a serious problem of representativeness.
If the relationship between age and emotional intelligence is a big new finding, it's worth doing a study with a broader age range. If, as I suspect, this is known anyway, you would be better advised not to present your finding on the grounds of the limited age distribution
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size