0
$\begingroup$

Assume that I have a data for the $ y $ like

          y
2000   74.5
2001   73.5
2002   71.4
2003   70.3
2004   79.1
...    ...

Also, I have data on $ x $.

           x
2000   123.5
2001   136.5
2002   243.4
2003   235.3
2004   278.1
...     ...

Can I simply transform this data into two variables for OLS estimation and just forget about time-component as follows

   y             x
74.5         123.5
73.5         136.5
71.4         243.4
70.3         235.3
79.1         278.1
 ...          ...

and estimate a model like this

$ y_i = \beta_0 + \beta_1 x_i + \epsilon_i $?

Will it be meaningful at all? Or why it can be wrong?


UPDATE: data to play with

$\endgroup$
1
$\begingroup$

No, you can't ignore time. You want to perform a transfer function model. You might have the need for differencing, ARIMA, denominator impacts on X transferring into Y. See Chapter 10 of the Box-Jenkins textbook. Post your data and we can dig deeper. Is there any lead or lag relationship between x and y?

$\endgroup$
  • $\begingroup$ @Tom_Reilly, thanks. Let's imagine that $ y $ is life expectancy in some country and $ x $ is a number of beds in hospitals per 10 000 people. I think the number of beds is lagged in comparison with life expectancy. $\endgroup$ – Vladimir Yashin Mar 10 '16 at 20:51
  • 1
    $\begingroup$ Take a look at the Cross Correlation Function(CCF) to see if there is a relationship. Post your data. You will also need a forecast of your causal. Do you have one already? $\endgroup$ – Tom Reilly Mar 12 '16 at 3:17
  • $\begingroup$ @Tom_Reilly, there is a lot to post. I provided an imaginary example for 2 of them, but I have more than 5 regressors. Thanks for the advice anyway. I will take a look. $\endgroup$ – Vladimir Yashin Mar 12 '16 at 3:51
  • $\begingroup$ Put it on dropbox.com. 5 Regressors is not a problem. Nor 50. $\endgroup$ – Tom Reilly Mar 12 '16 at 13:11
  • $\begingroup$ @Tom_Reilly, I added dataset. The first row contains headers, the second --- brief descriptions of a variables, the following --- data itself. I am trying to find a relationship between life expectancy (life_exp) and other ones. $\endgroup$ – Vladimir Yashin Mar 12 '16 at 20:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.