
123456

Thanks, on behalf of those who wanted it.
In terms of your original question I can see two situations here.
1. The data come as ranks. No comment to add.
2. The data do not come as ranks, but working with ranks is a reaction
to some queasiness about the data, e.g. outliers, skewness, some other
kind of messiness. Even in this situation I would tend to play a bit and
consider transformation vs no transformation, or glm with different
link functions. If the substantive conclusion is that the scale is
secondary, you can avoid KruskalWallis. If that the scale is crucial,
you (or your client) may need better data.
Nick
[hidden email]
Ricardo Ovaldia
Here it is:
Sokal, R. R. and F. J. Rohlf. 1995. Biometry: the principles and
practice of statistics in biological research. 3rd edition. W. H.
Freeman and Co.: New York.
>
> > From: Nick Cox < [hidden email]>
> > Date: Monday, March 15, 2010, 11:22 AM
> > . findit Kruskal
> >
> > turns up what there is. findit is a variant on
> search,
> > recommended
> > to you in the last thread you started.
> >
> > Minimal name (year) references are deprecated on this
> list.
> > But the
> > procedure recommended in that book is almost
> certainly
> > highly
> > programmable.
>
> Ricardo Ovaldia
>
> > Is there a PostHoc test for Kruskal Wallis in Stata?
> > Something like the one proposed by Sokal and Rohlf
> (1995).
*
* 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/


sysuse auto, clear
gen gpm = 1/mpg
qreg gpm weight
predict pred
predict se, stdp
local level = (100  95) / 200
gen ul = pred + invttail(e(df_r), `level') * se
gen ll = pred  invttail(e(df_r), `level') * se
twoway rarea ll ul weight, sort astyle(ci)  ///
line pred weight, sort  ///
scatter gpm weight, legend(off) ytitle(Gallons per mile)
Nick
[hidden email]
Jeffrey B. Wolpin
In Stata 10, assuming a qreg regression with one predictor, is there a
way to develop confidence intervals around the qreg trend line (or any
quantile for that matter)?
Also, once developed, what would be the syntax to then graph the scatter
plot, trend line and confidence intervals?
*
* 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/


Dear Mr. Cox,
Thank you. The example syntax worked great.
Question:
What is being computed by the line "local level = (100  95) / 200"?
Is this line in any way dependent upon the number of observations?
Alternatively, what does 200 signify.
Thanks again and pardon my ignorance.
Jeffrey B. Wolpin
Original Message
From: [hidden email]
[mailto: [hidden email]] On Behalf Of Nick Cox
Sent: Monday, March 15, 2010 1:41 PM
To: [hidden email]
Subject: RE: st: CIs around Qreg trend line
sysuse auto, clear
gen gpm = 1/mpg
qreg gpm weight
predict pred
predict se, stdp
local level = (100  95) / 200
gen ul = pred + invttail(e(df_r), `level') * se
gen ll = pred  invttail(e(df_r), `level') * se
twoway rarea ll ul weight, sort astyle(ci)  ///
line pred weight, sort  ///
scatter gpm weight, legend(off) ytitle(Gallons per mile)
Nick
[hidden email]
Jeffrey B. Wolpin
In Stata 10, assuming a qreg regression with one predictor, is there a
way to develop confidence intervals around the qreg trend line (or any
quantile for that matter)?
Also, once developed, what would be the syntax to then graph the scatter
plot, trend line and confidence intervals?
*
* 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/*
* 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/


<>
" What is being computed by the line "local level = (100  95) / 200"?"
Try Maarten`s http://www.statajournal.com/article.html?article=st0137 to
get a handle on this issue.
Preprint at
http://www.unituebingen.de/fileadmin/Uni_Tuebingen/Fakultaeten/SozialVerhalten/Institut_fuer_Soziologie/Dokumente/Pdf_Dateien/Buis/pvalue_pdf.pdf
HTH
Martin
Original Message
From: [hidden email]
[mailto: [hidden email]] On Behalf Of Jeff
Sent: Dienstag, 16. März 2010 00:27
To: [hidden email]
Subject: RE: st: CIs around Qreg trend line
Dear Mr. Cox,
Thank you. The example syntax worked great.
Question:
What is being computed by the line "local level = (100  95) / 200"?
Is this line in any way dependent upon the number of observations?
Alternatively, what does 200 signify.
Thanks again and pardon my ignorance.
Jeffrey B. Wolpin
Original Message
From: [hidden email]
[mailto: [hidden email]] On Behalf Of Nick Cox
Sent: Monday, March 15, 2010 1:41 PM
To: [hidden email]
Subject: RE: st: CIs around Qreg trend line
sysuse auto, clear
gen gpm = 1/mpg
qreg gpm weight
predict pred
predict se, stdp
local level = (100  95) / 200
gen ul = pred + invttail(e(df_r), `level') * se
gen ll = pred  invttail(e(df_r), `level') * se
twoway rarea ll ul weight, sort astyle(ci)  ///
line pred weight, sort  ///
scatter gpm weight, legend(off) ytitle(Gallons per mile)
Nick
[hidden email]
Jeffrey B. Wolpin
In Stata 10, assuming a qreg regression with one predictor, is there a
way to develop confidence intervals around the qreg trend line (or any
quantile for that matter)?
Also, once developed, what would be the syntax to then graph the scatter
plot, trend line and confidence intervals?
*
* 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/*
* 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/*
* 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/


Ricardo Ovaldia wrote:
Is there a PostHoc test for Kruskal Wallis in Stata?
Something like the one proposed by Sokal and Rohlf (1995).

I'm not sure whether it's the same test that you're referring to, but last
summer I had an occasion to program the Dwass, Steel, CritchlowFligner test. I
used M. Hollander & D. A. Wolfe, _Nonparametric Statistical Methods_ Second
Edition. (New York: John Wiley & Sons, 1999) pp. 24044 as the source for the
formula and for the worked example to verify the ado file's output.
It's not difficult to program, but I'll put it up on SSC once I write a help
file. You'll need to install John Gleason's qsturng from SSC in order to get
the reference distribution for the test statistic (Studentized range) if you
want to avoid manual lookup.
Joseph Coveney
*
* 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/


(100  95) / 200 is .025 as I presume you know or can work out for
yourself.
The result is put in a local macro called level which is then fed as
one argument to the invttail() function. In this case, you get a 95%
confidence interval. The other argument is the number of residual
degrees of freedom to play with. The element "df" is a clue here.
To get this example, I modified some code that you can see for a similar
problem in
. ssc type rcspline.ado
I use this form because it is then easy to see that for "different
values of 95", you would just plug in something else instead of 95, e.g.
local level = (100  90) / 200
is the different line needed for a 90% confidence interval.
Nick
[hidden email]
Jeffrey B. Wolpin
Thank you. The example syntax worked great.
Question:
What is being computed by the line "local level = (100  95) / 200"?
Is this line in any way dependent upon the number of observations?
Alternatively, what does 200 signify.
Nick Cox
sysuse auto, clear
gen gpm = 1/mpg
qreg gpm weight
predict pred
predict se, stdp
local level = (100  95) / 200
gen ul = pred + invttail(e(df_r), `level') * se
gen ll = pred  invttail(e(df_r), `level') * se
twoway rarea ll ul weight, sort astyle(ci)  ///
line pred weight, sort  ///
scatter gpm weight, legend(off) ytitle(Gallons per mile)
Jeffrey B. Wolpin
In Stata 10, assuming a qreg regression with one predictor, is there a
way to develop confidence intervals around the qreg trend line (or any
quantile for that matter)?
Also, once developed, what would be the syntax to then graph the scatter
plot, trend line and confidence intervals?
*
* 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/


Thank you Joseph. I am in the process of programing and testing Sokal and Rohlf which also uses the Q statistic. I am using John Gleason's qsturng command for the distribution. I would be intereted in seen if they are both the same method.
Regards,
Ricardo
Ricardo Ovaldia, MS
Statistician
Oklahoma City, OK
 On Tue, 3/16/10, Joseph Coveney < [hidden email]> wrote:
> From: Joseph Coveney < [hidden email]>
> Subject: st: Re: PostHoc test for Kruskal Wallis
> To: [hidden email]
> Date: Tuesday, March 16, 2010, 6:52 AM
> Ricardo Ovaldia wrote:
>
> Is there a PostHoc test for Kruskal Wallis in Stata?
> Something like the one proposed by Sokal and Rohlf (1995).
>
> 
>
> I'm not sure whether it's the same test that you're
> referring to, but last
> summer I had an occasion to program the Dwass, Steel,
> CritchlowFligner test. I
> used M. Hollander & D. A. Wolfe, _Nonparametric
> Statistical Methods_ Second
> Edition. (New York: John Wiley & Sons, 1999) pp.
> 24044 as the source for the
> formula and for the worked example to verify the ado file's
> output.
>
> It's not difficult to program, but I'll put it up on SSC
> once I write a help
> file. You'll need to install John Gleason's qsturng
> from SSC in order to get
> the reference distribution for the test statistic
> (Studentized range) if you
> want to avoid manual lookup.
>
> Joseph Coveney
>
>
> *
> * 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/>
*
* 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/


Thanks Nick and Jeph for your replies
Nick Cox wrote:
I think the main trick is to keep all the locals you need together with
the graph commands in a _small_ .do file. That is, as a .do file can
call other .do files, storing definitions together with commands is the
most obvious way to go.
I agree but in my work I have to do a lot of "mass production" where
the output is nearly the same for a lot of customers. So I often end up
with huge dofiles and loops. There might be other ways but despite the issue
discussed here I'm quite comfortable with that.
Jeph Herrin wrote a script to show that locals and contents of locals
can be parsed from macro list.
The idea works well but in this form only for simple locals.
I have some special cases so I altered it according to them:
There are issues with long locals going over several rows in macro dir.
I decided to take only the local names from local dir and take the
content right out of the local.
I often have several parts each containing several numbers or words
in compound quotes in one local (e.g. different graph titles).
After a linebreak in macro dir it could happen that something like
2007"'
was the first word and caused errors in the stringpos function
even if compound quotes in stringpos were used.
I eliminated all quotes from the first word of each line.
Here's the altered script:
It worked with all of my locals but I'm not sure if it can handle
each and every special case.
**********
capture log close
log using mylocals2.txt, text replace
macro list
log close
file open logfile using mylocals2.txt, text read
file open mylocals using mylocals2.do, text replace write
file read logfile line
while r(eof)==0 {
local lname : word 1 of `line'
local lname : subinstr local lname `"""' ""
local lname : subinstr local lname `"'"' ""
if strpos("`lname'","_")==1 {
local lval : subinstr local line "`lname'" ""
local lname : subinstr local lname "_" ""
local lname : subinstr local lname ":" ""
file write mylocals "local `lname' " _char(34) `"``lname''"' _char(34) _n
}
file read logfile line
}
file close logfile
file close mylocals
**********
Thanks again
Stefan
*
* 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/


I agree that "mark&find" is a nice feature but as far as I know
it's not documented (PDFdocumentation GSW[13] paragraph "searching").
Different software uses this in different ways so the user never knows.
One can argue that  after the search box is closed  the search ends
and F3 provides an additional search function based on the last inserted or marked string.
But on the other hand F3 is named "Find Next"
which makes it a logical successor of "Find".
In general it would be better not to have a search box but a search field
in the top or bottom of the editor GUI (e.g. like in Firefox or Acrobat reader)
which shows the active search string to the user and doesn't interfere with the text window.
Stefan
*
* 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/


<>
" but a search field
in the top or bottom of the editor GUI (e.g. like in Firefox or Acrobat reader)"
Or in the Viewer shipped with Stata  where the F3 key, in turn, seems not to be functional...
HTH
Martin
Ursprüngliche Nachricht
Von: [hidden email] [mailto: [hidden email]] Im Auftrag von [hidden email]
Gesendet: Mittwoch, 24. März 2010 10:55
An: [hidden email]
Betreff: st: Find Again and Copy in dofile editor
I agree that "mark&find" is a nice feature but as far as I know
it's not documented (PDFdocumentation GSW[13] paragraph "searching").
Different software uses this in different ways so the user never knows.
One can argue that  after the search box is closed  the search ends
and F3 provides an additional search function based on the last inserted or marked string.
But on the other hand F3 is named "Find Next"
which makes it a logical successor of "Find".
In general it would be better not to have a search box but a search field
in the top or bottom of the editor GUI (e.g. like in Firefox or Acrobat reader)
which shows the active search string to the user and doesn't interfere with the text window.
Stefan
*
* 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/*
* 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/


Dear Johan,
length of election period is an issue in this calculation.
If you have the length of the election period you can turn the proportional timevar into an absolute time var.
Perhaps you can take the length of the election period from the start and end dates of governments that didn't fail
(Assuming that election periods remain constant over time in one country).
That's the easy part.
Time to the next election seems to be quite of importance for government action so it might influence the failure risk years before the election period ends. So you probably have timeindependent risk factors (political events, affairs), timedependent risk factors from the start (wear, growing animosities in a coalition) and timedependent risk factors towards the end (tactical behavior of reelection favourite, ending of a stalemate situation).
So if one election period is four and another is five years it's probably not done only by censoring at four years.
Stefan
***********
Dear Statausers,
I have question regarding how to set up failuretime data in Stata.
In my dataset (an unbalanced panel of data on governments from 19452009) I have a proportional timevariable, i.e. the relative duration time measuring if government cabinets survives a whole term for which is elected, where 1 indicates that is stays in power for a whole term and e.g. 0.5 the cabinet dissolves after half a term. It is possible to use such a variable for setting up failuretime data in Stata?
I also have the start date and end date of cabinets, but since the timevariable above also contains qualitative judgements (of experts) what constitutes an failure, e.g. as some countries does not have fixed terms, it would be appropriate to use the relative duration variable described above.
Thanks,
Johan Hellström
*
* 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/


Dear Amy,
Stata date and time functions are somehow confusing at the start but very nice to work with later.
If you have the dates in Stata format (number of days from 1,1, 1960)
you don't have to care about the formats and just can calculate:
Var2date  Var1date = date difference.
You can multiply the date difference by 24*60 to change it into minutes.
********* Now to the time of day
I can't guess what your time vars contain.
You have a string format (str5) so you need to transfer the string vars into numeric vars first.
Example:
Replace Var1time_n = real(Var1time)
I suppose the vars contain elapsed seconds or minutes from midnight.
In this case things are rather easy:
You calculate the difference Var2time  Var1time which might be positive or negative.
Then convert this result into minutes if Var[1/2]time counts other time measures.
Finally add the time difference to the date difference (both in minutes) and you have the total difference.
This is one easy way, there are other functions available in the date and time section but one has to know the exact data specification to use them.
Stefan Gawrich
Dear Statalist users,
I am very much a beginner in Stata, and I have Stata 11.0. I have read the “help dates” instructional from Stata, but am still confused. For my project in the Zambian hospital setting, I must calculate the elapsed time in minutes of 2 sets of variables, they are:
Var1date: Date patient departed from the clinic
Var1time: Time patient departed from the clinic
Var2date: Date patient arrived at the hospital
Var2time: Time patient arrived at the hospital
The time variables have type “str5” and format “%5s.” The date variables have the type “long” and format “%dD_m_Y.” (I also do not quite understand the meaning of “str5,” “%5s,” and “long.”)
I would like to calculate the time difference (in minutes) between the time the patient left the clinic to when the patient arrived at the hospital. How would I merge, for example, Var2date and Var2time into “Var2” from which I may subtract Var1 (Var2Var1 = time transport takes).
Thanks very much for your consideration,
Amy
Dear Statalist users,
I am very much a beginner in Stata, and I have Stata 11.0. I have read the “help dates” instructional from Stata, but am still confused. For my project in the Zambian hospital setting, I must calculate the elapsed time in minutes of 2 sets of variables, they are:
Var1date: Date patient departed from the clinic
Var1time: Time patient departed from the clinic
Var2date: Date patient arrived at the hospital
Var2time: Time patient arrived at the hospital
The time variables have type “str5” and format “%5s.” The date variables have the type “long” and format “%dD_m_Y.” (I also do not quite understand the meaning of “str5,” “%5s,” and “long.”)
I would like to calculate the time difference (in minutes) between the time the patient left the clinic to when the patient arrived at the hospital. How would I merge, for example, Var2date and Var2time into “Var2” from which I may subtract Var1 (Var2Var1 = time transport takes).
Thanks very much for your consideration,
Amy
*
* 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/


<>
Stefan, how come your subject line suddenly changes to "st: bottom to top or reverse cumulative distribution in table command?" even though you are replying to Amy`s "Merging 2 variables to calculate time elapse (time/date)"? Is this a malfunction of the server?
BTW,  replace Var1time_n = real(Var1time) implies that "Var1time_n" has come into being somehow, maybe generated somewhere "upstream". The real() function will accept only strings that can be readily converted to numeric vars, w/o any removal of dashes, colons etc. So we do need to know what Amy`s data look like to give further advice...
***
di real("1231")
di real("1231")
di real("1231")
***
HTH
Martin
Ursprüngliche Nachricht
Von: [hidden email] [mailto: [hidden email]] Im Auftrag von [hidden email]
Gesendet: Donnerstag, 15. April 2010 12:26
An: [hidden email]
Betreff: AW: st: bottom to top or reverse cumulative distribution in table command?
Dear Amy,
Stata date and time functions are somehow confusing at the start but very nice to work with later.
If you have the dates in Stata format (number of days from 1,1, 1960)
you don't have to care about the formats and just can calculate:
Var2date  Var1date = date difference.
You can multiply the date difference by 24*60 to change it into minutes.
********* Now to the time of day
I can't guess what your time vars contain.
You have a string format (str5) so you need to transfer the string vars into numeric vars first.
Example:
Replace Var1time_n = real(Var1time)
I suppose the vars contain elapsed seconds or minutes from midnight.
In this case things are rather easy:
You calculate the difference Var2time  Var1time which might be positive or negative.
Then convert this result into minutes if Var[1/2]time counts other time measures.
Finally add the time difference to the date difference (both in minutes) and you have the total difference.
This is one easy way, there are other functions available in the date and time section but one has to know the exact data specification to use them.
Stefan Gawrich
Dear Statalist users,
I am very much a beginner in Stata, and I have Stata 11.0. I have read the “help dates” instructional from Stata, but am still confused. For my project in the Zambian hospital setting, I must calculate the elapsed time in minutes of 2 sets of variables, they are:
Var1date: Date patient departed from the clinic
Var1time: Time patient departed from the clinic
Var2date: Date patient arrived at the hospital
Var2time: Time patient arrived at the hospital
The time variables have type “str5” and format “%5s.” The date variables have the type “long” and format “%dD_m_Y.” (I also do not quite understand the meaning of “str5,” “%5s,” and “long.”)
I would like to calculate the time difference (in minutes) between the time the patient left the clinic to when the patient arrived at the hospital. How would I merge, for example, Var2date and Var2time into “Var2” from which I may subtract Var1 (Var2Var1 = time transport takes).
Thanks very much for your consideration,
Amy
Dear Statalist users,
I am very much a beginner in Stata, and I have Stata 11.0. I have read the “help dates” instructional from Stata, but am still confused. For my project in the Zambian hospital setting, I must calculate the elapsed time in minutes of 2 sets of variables, they are:
Var1date: Date patient departed from the clinic
Var1time: Time patient departed from the clinic
Var2date: Date patient arrived at the hospital
Var2time: Time patient arrived at the hospital
The time variables have type “str5” and format “%5s.” The date variables have the type “long” and format “%dD_m_Y.” (I also do not quite understand the meaning of “str5,” “%5s,” and “long.”)
I would like to calculate the time difference (in minutes) between the time the patient left the clinic to when the patient arrived at the hospital. How would I merge, for example, Var2date and Var2time into “Var2” from which I may subtract Var1 (Var2Var1 = time transport takes).
Thanks very much for your consideration,
Amy
*
* 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/*
* 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/


Dear Amy,
Stata date and time functions are somehow confusing at the start but very nice to work with later.
If you have the dates in Stata format (number of days from 1,1, 1960)
you don't have to care about the formats and just can calculate:
Var2date  Var1date = date difference.
You can multiply the date difference by 24*60 to change it into minutes.
********* Now to the time of day
I can't guess what your time vars contain.
You have a string format (str5) so you need to transfer the string vars into numeric vars first.
Example:
Replace Var1time_n = real(Var1time)
I suppose the vars contain elapsed seconds or minutes from midnight.
In this case things are rather easy:
You calculate the difference Var2time  Var1time which might be positive or negative.
Then convert this result into minutes if Var[1/2]time counts other time measures.
Finally add the time difference to the date difference (both in minutes) and you have the total difference.
This is one easy way, there are other functions available in the date and time section but one has to know the exact data specification to use them.
Stefan Gawrich
Dear Statalist users,
I am very much a beginner in Stata, and I have Stata 11.0. I have read the “help dates” instructional from Stata, but am still confused. For my project in the Zambian hospital setting, I must calculate the elapsed time in minutes of 2 sets of variables, they are:
Var1date: Date patient departed from the clinic
Var1time: Time patient departed from the clinic
Var2date: Date patient arrived at the hospital
Var2time: Time patient arrived at the hospital
The time variables have type “str5” and format “%5s.” The date variables have the type “long” and format “%dD_m_Y.” (I also do not quite understand the meaning of “str5,” “%5s,” and “long.”)
I would like to calculate the time difference (in minutes) between the time the patient left the clinic to when the patient arrived at the hospital. How would I merge, for example, Var2date and Var2time into “Var2” from which I may subtract Var1 (Var2Var1 = time transport takes).
Thanks very much for your consideration,
Amy
*
* 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/

123456
