[Carlo Lazzaro]

Georgios:
so you would probably be better off with switching to OLS.

[Georgios Kyrkos]

Mr. Lazzaro

There is a violation of OLS assumptions. The dependent variable is not normally distributed. So i suppose i need some non parametric models.


Best Regards

[Carlo Lazzaro]

Georgios:
it is not a violation, but a popular misconception that OLS prerequisites encompass the normal distribution of -depvar-.
Residuals should follow a normal distribution (but see for an interesting debate the following thread: http://www.statalist.org/forums/foru...ly-transformed).

[Georgios Kyrkos]

Dear Mr. Lazzaro

That is indeed an interesting debate. However, it is quite risky to treat the misconception as a scientific fact for my master thesis. Dont you think?
Unless if there is available literature.

[Carlo Lazzaro]

Georgios:
any decent textbook on econometrics can support that normality of The -depvar- is not a prerequisite for OLS..
Please, call me Carlo (as most on and off the list do).

[Georgios Kyrkos]

Dear Carlo

My apologies.
I am testing my dataset for autocorrelation. I am expecting to find for the variables Income and GDP. However when i use xtgls i received the following message:

 xtgls INC GDP
matsize too small - should be at least 328173
r(908);

The number 328173 is exactly the number of my observations. Does it have to do with the panel shape? (it is long)

Kind Regards

[Carlo Lazzaro]

Georgios:
no apologizes needed!
What if you type:

set matsize 328173

?
Anyway, I suspect that the outcome will take hours before appearing.

[Georgios Kyrkos]

Dear Carlo

I have STATA/IC 14.2

 set matsize 328173
matsize must be between 10 and 800
r(198);

[Carlo Lazzaro]

Georgios:
that's a problem.
Can't some colleague with a more powerful Stata 14.2 release lend you a hand?

[Georgios Kyrkos]

Hi Carlo

Sorry for the belated answer. Unfortunately there is not availability for an upgraded version.I tried the option:

et memory 20m
set memory ignored.
    Memory no longer needs to be set in modern Statas; memory adjustments are performed on the fly automatically.

But it seems that my version as well as my research is doomed.

I would like to ask you, which normality test would you recommend for my dataset which is 300.000 obs?

[Carlo Lazzaro]

Georgios:
if you refer to -depvar- normality, this is not an issue; hence, no need to test it.

[Georgios Kyrkos]

Hello Carlo.

I would like to test if there is Autocorrelation for the var Income which changes through the time. My dataset is a panel long.

My first result is

pwcorr INC L(1/3).INC

             |      INC    L.INC   L2.INC   L3.INC
-------------+------------------------------------
         INC |   1.0000
       L.INC |        .        .
      L2.INC |        .        .        .
      L3.INC |        .        .        .        .

Then i created 3 variables expressing the income per year (so L.INC1, L.INC2, LINC3)
Then my results are the following.

 pwcorr lag1 lag2 lag3

| lag1 lag2 lag3
-------------+---------------------------
lag1 | 1.0000
lag2 | . 1.0000
lag3 | . . 1.0000
]

.  pwcorr ID lag1 lag2 lag3

             |       ID     lag1     lag2     lag3
-------------+------------------------------------
          ID |   1.0000
        lag1 |  -0.0084   1.0000
        lag2 |  -0.0179        .   1.0000
        lag3 |  -0.0570        .        .   1.0000

Is it correct this apporach?

Thank you and Best Regards

George

[Zihao Chen]

Hi all,

I got a question when I doing the vector autoregression (VAR) in STATA. There is an error occurred which show repeated time values in sample r(451). And then I do the duplicates check as recommend in the following URL https://www.stata.com/support/faqs/d...d-time-values/. I can do the "tsset firm dm, monthly" command, but when I do the VAR, the error shows again. Can anyone help me with that?
[ATTACH=CONFIG]n1574647[/ATTACH]

[Zihao Chen]

1

[Shubham Sharma]

How to determine the optimum lag length in panel data?