stationarity

[frosoe]

please can anyone help me??

what stationarity means?? why do we use it and why is it usefulness??

thank you

[frosoe]

please help me because is for my economic dissertation and i have to explain it

thank you

[malisah]

One of the key underlying assumption of regression restricts it's usage to Normally distributed dependent variable. In simple terms, regression techniques can only be employed when variable of interest follows normal distribution (the bell shaped curve you see on cover of every statistics book). Now, a variable follows normal distribution when the chance of observing a value "too" far from the mean value is close to zero. But, some real life data, especially time-series data, fail to exhibit this nice property due to fluctualtion around the mean being unbounded or has some systematic trend. This unbounded nature of variablity in the data is commonly known as non-stationarity. So, for us to be able to use any time series technique (which are essentially advanced form of simple regression) we need to make it stationary. In lay man terms, we must perform certain operations on data so that excessive volatile nature of data is conatined and is made more suitable for any regression analysis. (source: http://www.linkedin.com/answers/managem ... 9-25283493).

[frosoe]

malisah thank you for the answer :) i am doing the perron test and all variables are non-stationary and after i am doing the augmented dickey fuller test and variables are stationary.doing the johansen cointegration method i have to use my stationary variables or my non stationary variables??

thank you again for the answer

[startz]

One of the key underlying assumption of regression restricts it's usage to Normally distributed dependent variable. In simple terms, regression techniques can only be employed when variable of interest follows normal distribution (the bell shaped curve you see on cover of every statistics book). Now, a variable follows normal distribution when the chance of observing a value "too" far from the mean value is close to zero....(source: http://www.linkedin.com/answers/managem ... 9-25283493).

This is just wrong. Nothing requires the dependent variable to be normal.

[EViews Glenn]

What's interesting is that this part of the answer was unrelated to the crux of the question which concerned stationarity. What's amazing is that it was "chosen" as the "best answer" on the original site *and* that there were better answers.