Lag Lenght calculatoin

[KloudMarta]

Hello my dears, I apologies for my below question, it seems beginner questions, actually I am new to Eviews 7 and I need a support from you.
My study is addressed the relationship between trade openness and gross fixed capital formation for the period (1990-2011). Dependent variable is Gross Fixed Capital Formation. Independent variables are Trade Openness, Inflation, Real Exchange Rate, Real Interest Rate, Saving, Political Stability. I applied unit root test (PP) to test stationary variables, after that I used VAR Estimation to determined Lag Length Criteria. Here is my problem, when I entered 4 or 3 Lag , it shows me the following letter " Insufficient number of observations for specified maximum lag length ". When I entered 2 lag, it shows me " Lag of non positive number ". But when I entered 1 Lag, it works.
So I hope you help me to find solution for my problem.
Thanks a lot.

[EViews Glenn]

What is the frequency of your data?
I can't quite tell from your post which dependent variables you've specified for the VAR.

[KloudMarta]

Annual data frequency.
Dependent variable is gross fixed capital formation.

[EViews Glenn]

I'm confused. Are you doing a VAR or unit root test? If the former, you'l need more than one dependent variable.

[sidra]

hello! what is the criteria of choosing lag length for ARCH-LM test and during applying augmented dickey fuler test?

[KloudMarta]

Dears..

EViews7 choose criteria and lag length automatically.
First, I apply unit root test (ADF, PP) at automatically selection (criteria and lags). After that, I use VAR estimation (with stationary variables) to determine what lags and criteria I should use.

Is that true? Is the step I follow true?

[trubador]

My study is addressed the relationship between trade openness and gross fixed capital formation for the period (1990-2011). Dependent variable is Gross Fixed Capital Formation. Independent variables are Trade Openness, Inflation, Real Exchange Rate, Real Interest Rate, Saving, Political Stability. I applied unit root test (PP) to test stationary variables, after that I used VAR Estimation to determined Lag Length Criteria. Here is my problem, when I entered 4 or 3 Lag , it shows me the following letter " Insufficient number of observations for specified maximum lag length ". When I entered 2 lag, it shows me " Lag of non positive number ". But when I entered 1 Lag, it works.

When you switch to VAR estimation, your independent variables also become dependents within a multivariate framework. Therefore, you end up with a 7-variable VAR, which requires 22 and 29 observations per equation to estimate the parameters. However, you'll only have 19 and 18 observations (due to adjustments after lag specifications).

[KloudMarta]

you are right..

That's what happened with me.

That's mean,VAR estimation don't work with me?????.
So, now what should I do? what estimation should I use?

Can you please help me.

[trubador]

Even if you use simple OLS, you'll still need more observations to obtain reliable results. So, if you do not want to give up any of your variables, then you should increase the sample size in any case. Using quarterly values might be a solution.

[KloudMarta]

Ok...

Is there any solution for small sample or observations?

[KloudMarta]

I mean is there any solution for my current small sample (from 1990-2011)?.

[EViews Glenn]

Fewer lags.

[KloudMarta]

Did you mean, when I apply unit root test I should use fewer lags instead of use automatically selection???

Or when I apply VAR estimation I should entered fewer lag? and even if I do it, is that true, or the true is more lags????

sorry for my questions, I use EViews7 for first time.

Thanks.

[EViews Gareth]

Just to be clear, your questions have nothing to do with EViews really. The issue of not having enough observations due to too many lags is purely econometric/mathematical. You'll have the same issue no matter what package you use to estimate it. What you're trying to do is mathematically impossible. The only solution is to try to do something else. Either get more observations, or use fewer lags in the VAR.

[KloudMarta]

Thanks a lot.