hypothesis testing for difference between two means

[jimjohn]

does anyone here know how to compute hypothesis tests/confidence intervals for the difference between two means in SPSS? For example, I am comparing the volume growth of my certificates with balance less than $1000 versus the volume growth with balances greater than $1000. I want to see if the difference in the two means is statistically significant. Any ideas? I don't think comparing means does it because it has something about grouping and factors. thanks!

[ViAnn Beadle]

Comparing means does it either via the usual analysis of variance or an
independent samples t-test. The group variable or factor is a variable with
2 categories corresponding to your balance split. However, when you do this
you lose a lot of information. Unless you have a reason to suspect that
$1000 is the magic threshold here, why not just look at the correlation of
growth and balance.

Also, while you're at it, scatterplot growth and balance to see if there is
any observable relationship between the two.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
jimjohn
Sent: Friday, May 08, 2009 2:32 PM
To: [hidden email]
Subject: hypothesis testing for difference between two means

does anyone here know how to compute hypothesis tests/confidence intervals
for the difference between two means in SPSS? For example, I am comparing
the volume growth of my certificates with balance less than $1000 versus the
volume growth with balances greater than $1000. I want to see if the
difference in the two means is statistically significant. Any ideas? I don't
think comparing means does it because it has something about grouping and
factors. thanks!
--
View this message in context:
http://www.nabble.com/hypothesis-testing-for-difference-between-two-means-tp
23452014p23452014.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
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

=====================
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

[Bob Schacht-3]

At 10:32 AM 5/8/2009, jimjohn wrote:
>does anyone here know how to compute hypothesis tests/confidence intervals
>for the difference between two means in SPSS? For example, I am comparing
>the volume growth of my certificates with balance less than $1000 versus the
>volume growth with balances greater than $1000. I want to see if the
>difference in the two means is statistically significant. Any ideas? I don't
>think comparing means does it because it has something about grouping and
>factors. thanks!
>--

I understood your question a little differently from Viann. I agree with
her that comparing mean balances would not make sense.
But I understand that you want to compare "mean volume growth," whatever
that is. So you use your Balance variable as your grouping variable, and
test the difference in mean volume growth. Do I understand that correctly?

Bob Schacht


Robert M. Schacht, Ph.D. <[hidden email]>
Pacific Basin Rehabilitation Research & Training Center
1268 Young Street, Suite #204
Research Center, University of Hawaii
Honolulu, HI 96814

=====================
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

[Johnny Amora]

At 10:32 AM 5/8/2009, jimjohn wrote:
>does anyone here know how to compute hypothesis tests/confidence intervals
>for the difference between two means in SPSS? For example, I am comparing
>the volume growth of my certificates with balance less than $1000 versus the
>volume growth with balances greater than $1000. I want to see if the
>difference in the two means is statistically significant. Any ideas? I don't
>think comparing means does it because it has something about grouping and
>factors. thanks!
>--

You can use the Independent Sample TTest procedure. You should have numeric variables for GROWTH and BALANCE where GROWTH as your test variable and BALANCE as your grouping variable.  The BALANCE should be coded for example as follows: 1= balance less than $1000 and 2 = greater than $1000.  Then run the synax:

 

T-TEST GROUPS=BALANCE(1 2)
  /MISSING=ANALYSIS
  /VARIABLES=GROWTH
  /CRITERIA=CI(.95).

The confidence interval (CI) can be changed to what you want, say .99.  Hope I understand correctly your need.

 

Cheers,

Johnny

---

Get connected with chat on network profile, blog, or any personal website!
Yahoo! allows you to IM with Pingbox. Check it out!

[Marta Garcia-Granero]

Hi

At 10:32 AM 5/8/2009, jimjohn wrote:

does anyone here know how to compute hypothesis tests/confidence intervals
for the difference between two means in SPSS? For example, I am comparing
the volume growth of my certificates with balance less than $1000 versus the
volume growth with balances greater than $1000. I want to see if the
difference in the two means is statistically significant. Any ideas? I don't
think comparing means does it because it has something about grouping and
factors. thanks!
--

IF I understand your question correctly, "balances" is not categorical, but quantitative too. the t test can be used, without having to dichotomize balances.

See this example, using the General Survey data that comes with SPSS. Suppose we want to compare number of children between those who have less than 12 years of education vs those who have 12 or more years. Instead of defining two groups, just use the cut-off:

GET FILE='U.S. General Social Survey.sav'.

T-TEST
  GROUPS = educ(12)
  /VARIABLES = childs.

Anyway, I also agree with ViAnn: you are throwing away a lot ofinformation. See "Altman DG, Royston P. (2006) The cost of dichotomising continuous variables. 332, 1080. "

HTH,
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

[jimjohn]

Thanks for all the responses guys. Johnny, I appreciate you explaining how the grouping variable works. Just one follow up question, I understand if I go to Compare Means -> Independent Test, then this will conduct a t-test to compare means. My sample size is about 36, I believe if the sample size is >30, a z test should be used. Any ideas if SPSS can do a Z-test too to compare means?

To the posters who posted before, thanks for the suggestions. I would normally do exactly what you guys recommended, but the circumstances are different here. Someone else has already performed an analysis claiming that there is a difference between the two means, that the <$1000 certificate customers are more sensitive to interest rate changes than the >$1000 certificate customers. Since this is not an intuitive conclusion, I wanted to run this test to see if the results are significant.

Johnny Amora wrote

At 10:32 AM 5/8/2009, jimjohn wrote:
>does anyone here know how to compute hypothesis tests/confidence intervals
>for the difference between two means in SPSS? For example, I am comparing
>the volume growth of my certificates with balance less than $1000 versus the
>volume growth with balances greater than $1000. I want to see if the
>difference in the two means is statistically significant. Any ideas? I don't
>think comparing means does it because it has something about grouping and
>factors. thanks!
>--

You can use the Independent Sample TTest procedure. You should have numeric variables for GROWTH and BALANCE where GROWTH as your test variable and BALANCE as your grouping variable.  The BALANCE should be coded for example as follows: 1= balance less than $1000 and 2 = greater than $1000.  Then run the synax:
 
T-TEST GROUPS=BALANCE(1 2)
  /MISSING=ANALYSIS
  /VARIABLES=GROWTH
  /CRITERIA=CI(.95).

The confidence interval (CI) can be changed to what you want, say .99.  Hope I understand correctly your need.
 
Cheers,
Johnny


      Yahoo! Mail Now Faster and Cleaner. Experience it at http://ph.mail.yahoo.com  today!

[Marta Garcia-Granero]

jimjohn wrote:
> Thanks for all the responses guys. Johnny, I appreciate you explaining how
> the grouping variable works. Just one follow up question, I understand if I
> go to Compare Means -> Independent Test, then this will conduct a t-test to
> compare means. My sample size is about 36, I believe if the sample size is
>
> 30, a z test should be used.
Wrong: If sample size  is >30, then a t test could (not should, notice
the difference) be replaced by a Z test. This is old statistical
thinking, from times when calculations were performed by hand, and
tricks to simplify them were common. The logic behind is that as sample
size increases, the t distribution tends asymptotically to a Z
distribution, therefore the error in the p-value is small under that
condition (big sample size). Today, since the hrd/boring job is done by
software, it is better to use a two samples t test and get the correct
p-value, not  an estimation with low error.
>  Any ideas if SPSS can do a Z-test too to compare means?
>

Chapter 5 of  Statistics at Square One (available online at:
http://bmj.bmjjournals.com/collections/statsbk/5.dtl ) describes the method.
My syntax file for the book, available at
http://gjyp.nl/marta/BMJ%20-%20Stats%20Square%20One.SPS , gives MATRIX
code get it, but, as I told you before, don't use it.

Marta

--
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