95% significance test
Locked
 
|
Hi Everyone,
I have interest in knowing if anyone has had any problem when calculating significance test with pivot tables. I have the following table data: New Starters 108 24.7% 155 32.2% Switched 2 0.5% 6 1.2% Continued 321 73.5% 316 65.6% Discontinued 6 1.4% 5 1.0% Total 437 482 but the comparison of columns will not show ANY significance at 95%. I've used the following link to check if spss was calculating the significance correct and I found that the are difference in "new starters". Could anyone give me a reason why spss is not showing up significance level at 95%? Thank in advance for your help. Mils
mils
|
Administrator
|
Maybe you should start by posting your syntax so people don't need to try to guess what you have tried!
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
|
|
Hi,
that's the syntax: * Custom Tables. CTABLES /VLABELS VARIABLES=subgroup7 subgroup DISPLAY=LABEL /TABLE subgroup7 [C] BY subgroup [C][COUNT F40.0, COLPCT.COUNT PCT40.1] /CATEGORIES VARIABLES=subgroup7 ORDER=A KEY=VALUE EMPTY=INCLUDE TOTAL=YES POSITION=AFTER /CATEGORIES VARIABLES=subgroup ORDER=A KEY=VALUE EMPTY=INCLUDE /COMPARETEST TYPE=PROP ALPHA=0.05 ADJUST=BONFERRONI ORIGIN=COLUMN INCLUDEMRSETS=YES CATEGORIES=ALLVISIBLE MERGE=NO /TITLES TITLE='Table 37:GLOBAL: Biologic population dynamics (last 3 month) - Wave 3 to Wave 6'. Subgroup: 1,2,3,4,5,6 Wave 1 (Apr-Jun 2010) Wave 2 (Oct-Dec 2010) Wave 3 (Apr-Jun 2011) Wave 4 (Oct-Dec 2011) Wave 5 (Apr-Jun 2012) Wave 6 (Oct-Dec 2012) Subgroup7: 1,2,3,4 New Starters Switched Continued Discontinued Both single select. Unfortunately, I can't provide data. Hope it helps, Mils
mils
|
|
Administrator
|
"ADJUST=BONFERRONI "
-- Without data (even aggregated) no further comment!
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
|
|
Wow, thanks David, I wasn't aware that the Bonferroni adjust will make such a difference.
Thanks, mils
mils
|
|
Administrator
|
Indeed!
In fact Bonferroni is very (the most) conservative approach to multiple comparisons. In your case where k=4 the effective level of significance for each comparison would be .05/4 = .0125 .
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
|
|
In reply to this post by David Marso
Mils,
You may be able to provide data if you replace all the variable names and labels with things like, "Variable 1", "Variable 2", etc. Best, -Vik On Jan 22, 2013, at 6:48 AM, David Marso wrote: > "ADJUST=*BONFERRONI *" > -- > Without data (even aggregated) no further comment! > > > mils wrote >> Hi, >> >> that's the syntax: >> >> * Custom Tables. >> CTABLES >> /VLABELS VARIABLES=subgroup7 subgroup DISPLAY=LABEL >> /TABLE subgroup7 [C] BY subgroup [C][COUNT F40.0, COLPCT.COUNT PCT40.1] >> /CATEGORIES VARIABLES=subgroup7 ORDER=A KEY=VALUE EMPTY=INCLUDE >> TOTAL=YES POSITION=AFTER >> /CATEGORIES VARIABLES=subgroup ORDER=A KEY=VALUE EMPTY=INCLUDE >> /COMPARETEST TYPE=PROP ALPHA=0.05 ADJUST=BONFERRONI ORIGIN=COLUMN >> INCLUDEMRSETS=YES CATEGORIES=ALLVISIBLE MERGE=NO >> /TITLES TITLE='Table 37:GLOBAL: Biologic population dynamics (last 3 >> month) - Wave 3 to Wave 6'. >> >> >> Subgroup: 1,2,3,4,5,6 >> Wave 1 (Apr-Jun 2010) >> Wave 2 (Oct-Dec 2010) >> Wave 3 (Apr-Jun 2011) >> Wave 4 (Oct-Dec 2011) >> Wave 5 (Apr-Jun 2012) >> Wave 6 (Oct-Dec 2012) >> >> Subgroup7: 1,2,3,4 >> New Starters >> Switched >> Continued >> Discontinued >> >> Both single select. Unfortunately, I can't provide data. >> >> Hope it helps, >> >> Mils > > > > > > ----- > Please reply to the list and not to my personal email. > Those desiring my consulting or training services please feel free to email me. > -- > View this message in context: http://spssx-discussion.1045642.n5.nabble.com/95-significance-test-tp5717562p5717570.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 |
|
|
In reply to this post by David Marso
Actually, Scheffe is even more conservative than Bonferroni since it controls for all possiblle comparisons, not just pairwise.
Dr. Paul R. Swank, Professor Health Promotion and Behavioral Sciences School of Public Health University of Texas Health Science Center Houston -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of David Marso Sent: Tuesday, January 22, 2013 10:28 AM To: [hidden email] Subject: Re: 95% significance test Indeed! In fact Bonferroni is very (the most) conservative approach to multiple comparisons. In your case where k=4 the effective level of significance for each comparison would be .05/4 = .0125 . mils wrote > Wow, thanks David, I wasn't aware that the Bonferroni adjust will make > such a difference. > > Thanks, > > mils ----- Please reply to the list and not to my personal email. Those desiring my consulting or training services please feel free to email me. -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/95-significance-test-tp5717562p5717574.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 |
Administrator
|
That's what I was thinking too. But then I tried the following example:
GET FILE='C:\SPSSdata\1991 U.S. General Social Survey.sav'. ONEWAY age BY region /MISSING ANALYSIS /POSTHOC=SCHEFFE BONFERRONI ALPHA(0.05). For every pair-wise contrast, the Scheffé p-value is lower than the Bonferroni p-value: .257 < .298; .152 < .157; .006 < .004. If Scheffé's method is more conservative, it should have higher p-values. But doing this example has also raised another question: Can SPSS even do complex contrasts using Scheffé's method? If so, how to do it is not immediately obvious. Checking the Scheffé box in the post-hoc dialog just gives all pair-wise contrasts.
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
|
Administrator
|
In reply to this post by David Marso
Actually, The divisor (according to the Algorithms is K *(K-1)/2) K=4: 12/2 = 6 so .05/6=.00833...
However I am at a loss to the true 'correct' applicability of statistical testing within CTABLES and stand by my assertion in another thread from earlier today. @ Paul: IIRC from my school days Sometimes Scheffe is more powerful/less conservative than Bonferroni. That is when there are MANY comparisons considered. See bottom of http://www.itl.nist.gov/div898/handbook/prc/section4/prc473.htm for example. I guess one could build out some simulations and graph the boundary cases. Maybe another day, or if this discussion gets too interesting ;-) --
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
Standardized vs. Adjusted Standardized Residuals for Statistically Significant Chi-Square
|
|
When trying to determine which groups are contributing to a significant overall chi-square test (for contingency tables that are larger than 2x2), I have read about using the Standardized residuals (i.e., Standardized residual values > 2). However, SPSS also has the option to give Adjusted Standardized residuals. I have tried reading up on the Adjusted Standardized residuals, but am not clear on when (of if) it is more appropriate to use the Standardized or Adjusted Standardized residuals to determine differences between groups. Any clarification or guidance the group can provide would be greatly appreciated.
Thank You, Todd Zoblotsky ===================== 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 |
Re: Standardized vs. Adjusted Standardized Residuals for Statistically Significant Chi-Square
|
|
Todd,
Adjusted Standardized Residuals (as defined by Haberman) are most like standardized residuals from regression as they have a Normal(0,1) distribution. I would use them wherever available.
If you are using these as surrogates to rank contributions of cells to Chi-Squared, either form does a decent job but the Adjusted Standardized Residuals are preferable as they correct for unequal table marginals.
... Mark Miller
On Tue, Jan 22, 2013 at 1:06 PM, Todd Alan Zoblotsky (tzbltsky) <[hidden email]> wrote: When trying to determine which groups are contributing to a significant overall chi-square test (for contingency tables that are larger than 2x2), I have read about using the Standardized residuals (i.e., Standardized residual values > 2). However, SPSS also has the option to give Adjusted Standardized residuals. I have tried reading up on the Adjusted Standardized residuals, but am not clear on when (of if) it is more appropriate to use the Standardized or Adjusted Standardized residuals to determine differences between groups. Any clarification or guidance the group can provide would be greatly appreciated. |
Re: Standardized vs. Adjusted Standardized Residuals for Statistically Significant Chi-Square
|
|
Thank you very much, Mark. Todd From: Mark Miller [mailto:[hidden email]]
Todd, Adjusted Standardized Residuals (as defined by Haberman) are most like standardized residuals from regression as they have a Normal(0,1) distribution. I would use them wherever available. If you are using these as surrogates to rank contributions of cells to Chi-Squared, either form does a decent job but the Adjusted Standardized Residuals are preferable as they correct for unequal table marginals. ... Mark Miller On Tue, Jan 22, 2013 at 1:06 PM, Todd Alan Zoblotsky (tzbltsky) <[hidden email]> wrote: When trying to determine which groups are contributing to a significant overall chi-square test (for contingency tables that are larger than 2x2), I have read about using the Standardized residuals (i.e., Standardized residual values >
2). However, SPSS also has the option to give Adjusted Standardized residuals. I have tried reading up on the Adjusted Standardized residuals, but am not clear on when (of if) it is more appropriate to use the Standardized or Adjusted Standardized residuals
to determine differences between groups. Any clarification or guidance the group can provide would be greatly appreciated. |
|
|
In reply to this post by David Marso
I have another question weighting related.
If we weight the data, how does SPSS runs the significance test? By using the weighted data? Meaning using weighted bases/counts and weighted percentages/proportions? Or by using unweighted bases/counts and weighted percentages/proportions? I know that Ctables allows unweighted counts, but not sure if that should solve the problem? I just wonder, if for example, we are looking at some data which has been weighted and we show weighted percentages and unweighted bases, can we use these to run significance test? Hope this make sense. Thanks in advance for your help. Mils.
mils
|
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |