The advantages of STATA over SPSS

[Wienta Diarsvitri]

Hello all,

I am an SPSS users. Some of my friends are choosing to leave SPSS and learn STATA. What are the advantages of STATA over SPSS?

Thank you

Wienta

---

Looking for the perfect gift? Give the gift of Flickr!

[Ruben Geert van den Berg]

Dear Wienta,
 
This link will bring you to a paper which attempts an honest comparison between SPSS, Stata and SAS.
 
http://www.ats.ucla.edu/stat/TechnicalReports/number1_editedFeb_2_2007/ucla_ATSstat_tr1_1.1_0207.pdf
 
HTH,

Ruben van den Berg

Methodologist

TNS NIPO

E: [hidden email]

P: [phone]

I: www.tns-nipo.com

 

---

Date: Tue, 19 Jan 2010 03:15:59 -0800
From: [hidden email]
Subject: The advantages of STATA over SPSS
To: [hidden email]

Hello all,

I am an SPSS users. Some of my friends are choosing to leave SPSS and learn STATA. What are the advantages of STATA over SPSS?

Thank you

Wienta

---

Looking for the perfect gift? Give the gift of Flickr!

---

Express yourself instantly with MSN Messenger! MSN Messenger

[Ruben Geert van den Berg]

Dear all,
 
I sometimes encounter data in which effects are partly between and partly within subjects. I pasted a simplified example of this structure below. Cases with the same value on ID are a single person. Now if I'd like to run a (full factorial) ANOVA on this (meal cook meal*cook with rating as dependent), then which procedure should I use? Because every ID occurs twice, the assumption of independent observations is severely violated but if I restructure the data, it will be full of missing values. Is there any correct way to analyze this?
 
TIA!
 

data list free/ID Meal Cook Rating.
begin data
1 0 0 7
1 1 1 6
2 1 0 7
2 0 1 6
3 0 0 7
3 1 1 6
4 1 0 9
4 0 1 6
5 0 0 9
5 1 1 8
6 1 0 8
6 0 1 4
7 0 0 8
7 1 1 6
8 1 0 9
8 0 1 5
9 0 0 7
9 1 1 6
10 1 0 8
10 0 1 5
11 0 0 6
11 1 1 8
12 1 0 9
12 0 1 6
end data.
for all(f1).
val lab Meal 0'Steak'1'Lasagna'/Cook 0'Jamie Oliver'1'Gordon Ramsay'/Rating 0'Extremely bad'10'Extremely tasty'.
 
sor cas ID Meal cook.
casest
/id id
/ind meal cook
/gro var.

Ruben van den Berg

Methodologist

TNS NIPO

E: [hidden email]

P: [phone]

I: www.tns-nipo.com

 

---

New Windows 7: Simplify what you do everyday. Find the right PC for you.

[Steve Simon, P.Mean Consulting]

Wienta Diarsvitri wrote:

> I am an SPSS users. Some of my friends are choosing to leave SPSS and
> learn STATA. What are the advantages of STATA over SPSS?

You'll get a bunch of conflicting opinions, so don't take any single
opinion too seriously including mine. What package you use and like
often depends strongly on what package you used first in graduate school
or what package all your colleagues use.

I've used both packages in my previous job, but mostly SPSS. I reserved
Stata for the complex analyses that could not be done by SPSS. We had a
site license for SPSS, but not for Stata, so I could encourage my
clients to try SPSS on their own, but not Stata.

Stata is harder to learn than SPSS because it is predominantly a command
based interface with menu selections laid on top. SPSS is predominantly
a menu based interface with command selections (syntax) laid on top.

Stata is not that much more complicated, however. If you are already
comfortable with SPSS syntax, you will not have too much trouble with Stata.

Also, Stata is a single package and not a bunch of modules that you
choose from. That can be perceived as an advantage or as a disadvantage.

Stata is pretty easy to extend, so like R it has a lot of user
contributed content. I've used a user contributed module of Stata to
perform meta-analysis for example.

SPSS does have extensibility now, but it was later in adding it than
Stata, so there is less user created content, though Raynald Levesque is
doing a heroic job at www.spsstools.net.

I would argue that any statistician should have the ability to run
several different software packages. So I would encourage you to
consider SPSS and Stata or SPSS and R or SPSS and SAS. That can get
expensive, of course, because all the packages except for R cost a lot
of money. If your employer picks up the bill, that's less of a
consideration, but for an independent consultant like me, money is a
serious consideration.

I hope this helps.
--
Steve Simon, Standard Disclaimer
"The first three steps in a descriptive
data analysis, with examples in PASW/SPSS"
Thursday, January 21, 2010, 11am-noon, CST.
Free to all! Details at www.pmean.com/webinars

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Ruben Geert van den Berg]

Dear all,
 
I posted the question below last week but -so far- without any response. I'm not a big fan of posting the same question twice but for this one I thought I could give it a second try. Any advice/reference is more than welcome!
 
I sometimes encounter data in which effects are partly between and partly within subjects. I pasted a simplified example of this structure below. Cases with the same value on ID are a single person. Now if I'd like to run a (full factorial) ANOVA on this (meal cook meal*cook with rating as dependent), then which procedure should I use? Because every ID occurs twice, the assumption of independent observations is severely violated but if I restructure the data, it will be full of missing values. Is there any correct way to analyze this?
 
TIA!!
 
Ruben
 
data list free/ID Meal Cook Rating.
begin data
1 0 0 7
1 1 1 6
2 1 0 7
2 0 1 6
3 0 0 7
3 1 1 6
4 1 0 9
4 0 1 6
5 0 0 9
5 1 1 8
6 1 0 8
6 0 1 4
7 0 0 8
7 1 1 6
8 1 0 9
8 0 1 5
9 0 0 7
9 1 1 6
10 1 0 8
10 0 1 5
11 0 0 6
11 1 1 8
12 1 0 9
12 0 1 6
end data.
for all(f1).
val lab Meal 0'Steak'1'Lasagna'/Cook 0'Jamie Oliver'1'Gordon Ramsay'/Rating 0'Extremely bad'10'Extremely tasty'.
 
sor cas ID Meal cook.
casest
/id id
/ind meal cook
/gro var.

---

Express yourself instantly with MSN Messenger! MSN Messenger

[Marta Garcia-Granero]

Ruben van den Berg wrote:
>
> I posted the question below last week but -so far- without any
> response. I'm not a big fan of posting the same question twice but for
> this one I thought I could give it a second try. Any advice/reference
> is more than welcome!
I must have missed your first post...

>
> I sometimes encounter data in which effects are partly between and
> partly within subjects. I pasted a simplified example of this
> structure below. Cases with the same value on ID are a single person.
> Now if I'd like to run a (full factorial) ANOVA on this (meal cook
> meal*cook with rating as dependent), then which procedure should I
> use? Because every ID occurs twice, the assumption of independent
> observations is severely violated but if I restructure the data, it
> will be full of missing values. Is there any correct way to analyze this?
>
>

Try Googling "Incomplete balanced blocks design". You could treat ID
(persons) as blocks (not absolutely correct but...), combine Meal and
Cook into a single grouping variable (with values Grp1.1, Grp1.2, Grp2.1
& Grp2.2) and use a set of user defined contrasts to study main effects
& interaction (Ching Chun Li's way of working with 2x2 factorial
experiments with blocks):

Meal: -1 -1 1 1
Cook: -1 1 -1 1
M*C Int: 1 -1 -1 1

Don't restructure the dataset, use it as it is, select type I sum of
squares, and make sure that blocks (ID variable in your case) goes first
in /DESIGN.

I think it should work.

HTH,
Marta GG

>
> data list free/ID Meal Cook Rating.
> begin data
> 1 0 0 7
> 1 1 1 6
> 2 1 0 7
> 2 0 1 6
> 3 0 0 7
> 3 1 1 6
> 4 1 0 9
> 4 0 1 6
> 5 0 0 9
> 5 1 1 8
> 6 1 0 8
> 6 0 1 4
> 7 0 0 8
> 7 1 1 6
> 8 1 0 9
> 8 0 1 5
> 9 0 0 7
> 9 1 1 6
> 10 1 0 8
> 10 0 1 5
> 11 0 0 6
> 11 1 1 8
> 12 1 0 9
> 12 0 1 6
> end data.
> for all(f1).
> val lab Meal 0'Steak'1'Lasagna'/Cook 0'Jamie Oliver'1'Gordon
> Ramsay'/Rating 0'Extremely bad'10'Extremely tasty'.
>

--
For miscellaneous SPSS related statistical stuff, visit:
http://gjyp.nl/marta/

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Marta Garcia-Granero]

Hi again Ruben:

I spoke/wrote too fast. I have tried UNIANOVA-ing your sample dataset,
and, since Group has only 2 DF, I can only study main effects, because
the interaction is confounded ("melted") with the blocks. Google for
"Confounding factorial designs" to find more info.

Analysis of your sample dataset:

data list free/ID Meal Cook Rating.
begin data
1 0 0 7
1 1 1 6
2 1 0 7
2 0 1 6
3 0 0 7
3 1 1 6
4 1 0 9
4 0 1 6
5 0 0 9
5 1 1 8
6 1 0 8
6 0 1 4
7 0 0 8
7 1 1 6
8 1 0 9
8 0 1 5
9 0 0 7
9 1 1 6
10 1 0 8
10 0 1 5
11 0 0 6
11 1 1 8
12 1 0 9
12 0 1 6
end data.
for all(f1).
val lab Meal 0'Steak'1'Lasagna'/Cook 0'Jamie Oliver'1'Gordon
Ramsay'/Rating 0'Extremely bad'10'Extremely tasty'.

NUMERIC Group (F8).
IF (Meal=0) & (Cook=0) Group=1.
IF (Meal=0) & (Cook=1) Group=2.
IF (Meal=1) & (Cook=0) Group=3.
IF (Meal=1) & (Cook=1) Group=4.
VALUE LABEL Group 1'Steak/Oliver' 2'Steak/Ramsay' 3'Lasagna/Oliver'
4'Lasagna/Ramsay'.

* Way #1 *.
UNIANOVA
  Rating  BY Group ID
  /RANDOM = ID
  /METHOD = SSTYPE(1)
  /INTERCEPT = INCLUDE
  /PRINT = DESCRIPTIVE
  /CONTRAST (Group)=SPECIAL(-1 -1  1 1
                            -1  1 -1 1)
  /DESIGN = ID Group .

* Way #2 (same results as above) *.

UNIANOVA
  Rating  BY Meal Cook ID
  /RANDOM = ID
  /METHOD = SSTYPE(1)
  /INTERCEPT = INCLUDE
  /EMMEANS = TABLES(Meal)
  /EMMEANS = TABLES(Cook)
  /CRITERIA = ALPHA(.05)
  /DESIGN = ID Meal Cook .

Conclusion: your design has eliminated any chance of studying the
interaction. Was it important?

Best regards,
Marta GG

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

Ruben van den Berg wrote

Dear all,

I posted the question below last week but -so far- without any response. I'm not a big fan of posting the same question twice but for this one I thought I could give it a second try. Any advice/reference is more than welcome!

I sometimes encounter data in which effects are partly between and partly within subjects. I pasted a simplified example of this structure below. Cases with the same value on ID are a single person. Now if I'd like to run a (full factorial) ANOVA on this (meal cook meal*cook with rating as dependent), then which procedure should I use? Because every ID occurs twice, the assumption of independent observations is severely violated but if I restructure the data, it will be full of missing values. Is there any correct way to analyze this?

TIA!!

Ruben
 
data list free/ID Meal Cook Rating.
begin data
1 0 0 7
1 1 1 6
2 1 0 7
2 0 1 6
3 0 0 7
3 1 1 6
4 1 0 9
4 0 1 6
5 0 0 9
5 1 1 8
6 1 0 8
6 0 1 4
7 0 0 8
7 1 1 6
8 1 0 9
8 0 1 5
9 0 0 7
9 1 1 6
10 1 0 8
10 0 1 5
11 0 0 6
11 1 1 8
12 1 0 9
12 0 1 6
end data.
for all(f1).
val lab Meal 0'Steak'1'Lasagna'/Cook 0'Jamie Oliver'1'Gordon Ramsay'/Rating 0'Extremely bad'10'Extremely tasty'.
 
sor cas ID Meal cook.

HI Ruben.  I think I'd be looking into whether this can be done via MIXED.  The command syntax reference manual has this example of how to do a one-way repeated measures ANOVA via MIXED:

MIXED
  distance  BY age
  /FIXED = age  | SSTYPE(3)
  /REPEATED = age | SUBJECT(subject) COVTYPE(cs)
  /EMMEANS = TABLES(age)
.

Using that as a model, you might want to try something like the following:

* Main effects only model .

MIXED Rating BY Meal Cook
  /FIXED=Meal Cook | SSTYPE(3)
  /REPEATED = Meal Cook | SUBJECT(id) COVTYPE(cs)
  /METHOD=ML
  /PRINT=SOLUTION
  /EMMEANS=TABLES(Meal)
  /EMMEANS=TABLES(Cook)
.

* With the interaction is included--not sure if it will work! .

MIXED Rating BY Meal Cook
  /FIXED=Meal Cook Meal*Cook | SSTYPE(3)
  /REPEATED = Meal Cook | SUBJECT(id) COVTYPE(cs)
  /METHOD=ML
  /PRINT=SOLUTION
  /EMMEANS=TABLES(Meal)
  /EMMEANS=TABLES(Meal*Cook)
.

Good luck!

[Jennifer Day-2]

Hello out there!

Does anyone know if there is a command in Stata to write the results
of an analysis to an Excel file?  I don't want to export the database
(as 'outfile' would do).  Instead, I've generated some tables that are
large (too large to view all at once on the stata output screen).  I'd
like to write these into an excel file so I can play around with the
results.

I know of the command "outreg", which outputs regression results.  So
I know such a thing is possible.  But I can't figure out how to do
it!!!  So frustrating!  I've also tried using "log", but that outputs
results to an .smcl file -- not formatted for use in Excel.

Any ideas?

Thanks,
jennie

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD