Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Log Normality of Dependentvar


From   sjsamuels@gmail.com
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Log Normality of Dependentvar
Date   Tue, 9 Jun 2009 15:19:57 -0400

--
I queried Stata Technical support about how the "lnormal" option works
in -swilk-.  Wes Eddings sent me the following reply, slightly edited.

Steve
_____________________________________________
"The -lnnormal- option expects
the variable to have already been transformed.  The help file for the original
user-written -swilk- reads:

"If the -lnnormal- option is specified, the data are tested under the assumption
that they are of the form log(X-k), where k is a constant determined from the
data.  The data should be supplied already transformed to log(X-k)."

.  -swilk- does not call -lnskew0- because
-swilk- assumes that the data have already been transformed.

I have submitted a request to clarify the -swilk- help file and documentation."

______________________________________________

On Mon, Jun 8, 2009 at 12:38 PM, <sjsamuels@gmail.com> wrote:
> -Chris--
>
> -lnskew0-- finds  by iteration a value of k for which y= ln(x - k) has
> skewness zero.  The manual implies that with the "lnnormal" option,
> -swilk- , estimates "k" by the method of -lnskew0-.  In fact, the ado
> file for -swilk- does not call -lnskew0-, but instead computes an
> approximation.. This probably accounts for the discrepancy that you
> observed.
>
> Analyses of  ln(var) and of the transformation  -bcskew0- are
> irrelevant to -swilk-, because the 'lnnormal" option considers the
> hypothesis of a three-parameter lognormal distribution.   I presume
> that by "skskew0"  you meant  "lnskew0
>
> -Steve
>
>>
>> --- On Mon, 8/6/09, Christian Weiss wrote:
>>> testing my dependent var via swilk or sfrancia rejects the
>>> Null Hypothesis of Normality.
>>
>> This is problematic for a number of reasons:
>>
>> 1) Regression never assumes that the dependent variable is
>> normally distributed, except when you have no explanatory
>> variables. It only assumes that the residuals are normally
>> distributed.
>>
>> 2) Testing for the normality of the residuals should only
>> be done once you are confinced that the other assumptions
>> have been met, as violations of the other assumptions are
>> likely to lead to residuals that look non-normal
>>
>> 3) The normality of the residuals is probably the least
>> important of the regression assumptions, as regression
>> is reasonably robust to violations of it.
>>
>> 4) Tests are probably not the best way to assess whether
>> the errors are normaly distributed. Graphical inspection
>> is usually more informative and powerful, see:
>> -help diagnostic plots- and -ssc d hangroot- for tools
>> to help with that.
>>
>> For a more general set of tools to perform post-estimation
>> checks of  regression assumptions see:
>> -help regress postestimation-.
>>
>>
>
> On Mon, Jun 8, 2009 at 5:38 AM, Christian
> Weiss<christian.weiss@nightberry.de> wrote:
>>
>> testing my dependent var via swilk or sfrancia rejects the Null
>> Hypothesis of Normality.
>> However, using the "lnnormal" option of swilk accepts the nully
>> hypothesis - it seems that the dependent variable is lognormal
>> distributed.
>>
>>
>> Suprisingly,after transformim my dependent variable by ln(var) or by
>> skskew0 / bcskew0, swilk still rejects the null hypothesis of
>> normality.
>>
>> How can that be explained?
>>
>> ..puzzled...Chris
>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index