non-stationary data to stationary

[pfeffer]

Hi there,

I have some data which is the daily return from the S&P500 index calculated from the 1% most extreme increase in the price. I caculated the return by taking Pt=(pt/pt-1)-1. I also calculated data from the 2,5% extreme and 5%. For the most I can do an ADF and have the data as stationary and there after do a regression analysis. Although on the odd ocassion I get an ADF saying non-stationary with 50% probability. The questions is if it is possible to make the non-stationary data stationary without getting inferior data. I've tried to log the data with bad results as it doesn't make it stationary. So my question is whether it's possible to make this non-stationary data stationary somehow without, getting inferior data and results. Or it's just better to present the data as non-stationary and therefore no regression analysis is done hence spurious?

Thanks in advance

Simon

[BobJ]

I've never heard on non-stationary asset returns. You must be interpreting/performing the ADF in the wrong way. Returns are just like first differincing. They would make the price series stationary. Log differencing the price series is another option (you get continuously compounded returns). It is not advisable to run a regression with non-stationary variables, you can get spurious results.

[Mimo]

Hello,

i am trying to learn how to use Eviews and along with it also statistics.

I am currently trying to convert the non-stationary data to stationary, but after i go through the process, then dhe result is the variable DY (new seriesm, of 1st difference) with NA values. What i am not getting is, why there are no values in the differenced variable?

Can anyone maybe help me here?

Best,

[EViews Glenn]

There are many things that could be happening but without seeing your data and the exact steps you are doing it is difficult to say. Could you post your data here and describe what you did?

[Mimo]

100
96.43337420992959
96.22840639276665
96.69729349595629
100.4905191547895
105.7859339235774
110.9855525554616
115.8702886041815
130.7678366895695
126.3059164280364
114.0432058948166
100.9337491586102
93.44226832702295
101.0003378975999
96.0896144891812
86.15804649363639
82.27965495226396
84.38417697990551
84.61395560434235
89.66375989413505
94.49600519169238
90.45617827287471
87.37786302163845
92.54786143019649
106.9217817627356
114.9718773008566
114.9718773008566
96.85716013683512
98.44835440566249
103.0435553515281
105.2704801138903
105.2704801138903


After Unit Root Test,i have to take the 1st difference to turn the data into stationary. After i get the 1st difference, i want to genr series by equation, and what i get is: :series: with NA values.

The same happens when i try to write down the genr equation on the top bar of the workfile, and try to pres enter so that the new genr series shows up in the workfile, but nothing happens. How can i save the new data of 1st or 2nd difference in my workfile?
The data i am using are normalized (index 1980=100). Do i maybe have to use absolute values instead?

Related to this again, i am a bit confused with the stationarity and seasonality.
Which method from Eviews 7 is best to check for both stationarity and seasonality, and which method is best to deseasonalize and stationarize (if i may name it like this)?

''I think i have entered a forest without an ax'' :?

[startz]

Try

Code: Select all

temp = d(used_extraction) used_extraction = used_extraction - temp

Although I believe what you are computing is

Code: Select all

used_extraction = used_extraction(-1)

which is probably not what you intend.

[Mimo]

thank you so much startz.
Because i already lost the used extraction values (have only NA values now), i tried with another variable (used/cap) and the result i got is: check file attached.

Does this mean that i have to do the same for all my variables one by one, until i get the stationary versions, and then can continue with analysis and model development?

[startz]

I'm really not sure what you're trying to do, but if you want to first difference a variable X use

Code: Select all

series firstDifferenceX = D(X)

and then work with the variable firstDifferenceX without further adjustments.

[Mimo]

all i am trying to do is: converting all variables into stationary, as all of them are non-stationary. Some of them require 2nd difference, even.
i will try to find it also in the eviews users guide. ;-)
thanks again!
p.s. my lucky time to learn more about eviews, statistics and econometrics. Huhhhhh

[Vivianlfzhang]

I have a similar question for non-stationary data transformation. I do not understand why temp=d(x) and then x=x-temp. I thought d(x) returns the first difference of x so why bother to subtract it from x? Thanks

[Qurat]

i am working on the impact of exchange rate on returns. for returns i took stock prices and convert them to return by applying formula = LN(pt/pt-1). than i took these returns to eviews and apply ADF at level and my series were found stationary at level. now plz help me whether to treat it stationary at level or at 1st difference. i am confuse because by taking returns we are converting data at 1st difference.