0

I use QQ-plot on residuals in time series analysis and find the distribution is not normally distributed, but somehow thin-tail distributed:

enter image description here

I know that if the residuals are not a normal distribution, the model assumption does not hold and needs to transform the data to normality then fit the model. However, I searched some answers (how-to-transform-leptokurtic-distribution-to-normality) which are all related to transforming the heavy-tail distributed data to normal distribution.

My question is:
1.Can I use the light-tailed distribution to analyze data even under the normality assumption?
2.Does the heavy-tailed problem more severe than light-tailed problem?

Chen
  • 1
  • 1
    Whether a normality assumption is important depends on what you are actually doing. – Henry Apr 12 '21 at 08:47
  • For most models, such residuals would be considered even *better* than Normally distributed ones! (I strongly doubt these are real data, though. Residuals just don't behave like this.) – whuber Apr 12 '21 at 14:58
  • Are the data U(0,105)? (Not sure about the upper bound). Sure looks that way! – BigBendRegion Apr 13 '21 at 14:28

0 Answers0