16 Nov Interpreting Descriptive Statistics
While working on student papers, student thesis or research papers, most of the student find difficulty in finding ways to interpret the data using the table. In this blog, econistics will provide your some insights what can be done with the descriptive statistics table using examples.
Consider the example of descriptive statistics table for cross sectional or time series data.
First thing we can interpret is the relative dispersion of the data, if the mean is above standard deviation the data is under dispersed indicating that the scatter plots are closely located near to its mean value and vice versa.
Second thing is the presence of outliers, if the kurtosis value is located near 3, we can assume that there are neither too few nor too many outliers present in that particular data. If it is above 3 we should be careful in estimating the model as it might be leading to heteroscedastic estimates.
Since most of the variables in economics do not have its ideal (subjectively desirable) value in the middle hence we cannot expect any of the data sets to have no skewness. Hence you must ensure sample to be above 30 in order for the data to become asymptotically normal (based on central limit theorem).
You can further add probability values of Skewness, Kurtosis and Jarque Bera where you can assess hypothesis regarding the normality of the data.
Also most of the studies ignore to add growth rate of the data, this can be very useful indicator to assess how much data has changed especially in time series case. while for the case of cross sectional data, we can get help from range of the data.
Lastly for the case of panel data, you can add cross section and time series wise means and growth rates, which be estimated using xtsum command in Stata.
One objective of doing descriptive analysis is to get the reader aware about the circumstances of the data. Second using this table most of the estimation models can be decided. Few detailed discussions are provided here for time series and for cross section data sets.
Please provide your comments, we will be adding more cases based on suggestions provided by our followers.