Significant Chi sq but no group differences

[Tom]

Hello

 

Using a syntax from Marta GG, I made several comparisons that help me to decide, which “Gruppe” differ in the proportion of “Yes” significantly from each other. It seemed to me to working fine – until now with the data below.

 

What I’ve done and suppose:

I got a significant chi sq and an adjusted residual for the “Gruppe” 1-3 of 2 resp. 2.1 for 4-10, which tells me, I suppose, that in these two groups the proportion of “Yes” is significantly higher than overall. So, the following syntax should show me, if these two groups (or just one) have a significantly higher proportion of “Yes” compared two each of the other groups.

 

What I do not understand:

But, with these data, there is no significant difference in the proportion between the groups, contrary to what I expected, because according to my assumption the significant Chi sq means, that there follows at least one comparison that is significant.  

 

Thanks for any explanations.

Tom

 

 

*****Beurteilung Fremdsprachen*****.

DATA LIST LIST/Gruppe Result Count(3 F8).

BEGIN DATA

1 1 53

2 1 154

3 1 262

4 1 288

5 1 187

1 2 19

2 2 46

3 2 49

4 2 54

5 2 36

END DATA.

WEIGHT BY Count .

VALUE LABELs Result 1'No' 2 'Yes'.

VALUE LABELs Gruppe 1'1-3' 2'4-10' 3'11-20' 4'21-30' 5'31-40'.

* A) Using CTABLES with Bonferroni adjustment *.

CTABLES

  /VLABELS VARIABLES=Gruppe Result DISPLAY=DEFAULT

  /TABLE Result [COUNT F40.0, COLPCT.COUNT PCT40.1] BY Gruppe

  /SLABELS POSITION=ROW

  /CATEGORIES VARIABLES=Gruppe Result ORDER=A KEY=VALUE EMPTY=INCLUDE

  /COMPARETEST TYPE=PROP ALPHA=0.05 ADJUST=BONFERRONI ORIGIN=COLUMN

  INCLUDEMRSETS=YES CATEGORIES=ALLVISIBLE.

 

 

 

[kelvidpang]

Hello,

CROSSTABS
  /TABLES=Result BY Gruppe
  /FORMAT=AVALUE TABLES
  /STATISTICS=CHISQ
  /CELLS=COUNT COLUMN
  /COUNT ROUND CELL.

I'm using crosstab to check the significance of chi-square. In my opinion, the proportion of Yes for group 1-3 is 26.4% has dropped to group 31-40 which is 16.1%. The difference is about 10%, which I think the difference is great enough.

[Marta Garcia-Granero]

Hi Tom:

Didn't you get my reply to your previous post concerning lack of significant results using Bonferroni adjustment? Have you tried "ADJUST=NONE" as I said?

Best regards,
Marta Garcia-Granero

El 17/11/2010 12:09, Balmer Thomas escribió:

Hello

 

Using a syntax from Marta GG, I made several comparisons that help me to decide, which “Gruppe” differ in the proportion of “Yes” significantly from each other. It seemed to me to working fine – until now with the data below.

 

What I’ve done and suppose:

I got a significant chi sq and an adjusted residual for the “Gruppe” 1-3 of 2 resp. 2.1 for 4-10, which tells me, I suppose, that in these two groups the proportion of “Yes” is significantly higher than overall. So, the following syntax should show me, if these two groups (or just one) have a significantly higher proportion of “Yes” compared two each of the other groups.

 

What I do not understand:

But, with these data, there is no significant difference in the proportion between the groups, contrary to what I expected, because according to my assumption the significant Chi sq means, that there follows at least one comparison that is significant.  

 

Thanks for any explanations.

Tom

 

 

*****Beurteilung Fremdsprachen*****.

DATA LIST LIST/Gruppe Result Count(3 F8).

BEGIN DATA

1 1 53

2 1 154

3 1 262

4 1 288

5 1 187

1 2 19

2 2 46

3 2 49

4 2 54

5 2 36

END DATA.

WEIGHT BY Count .

VALUE LABELs Result 1'No' 2 'Yes'.

VALUE LABELs Gruppe 1'1-3' 2'4-10' 3'11-20' 4'21-30' 5'31-40'.

* A) Using CTABLES with Bonferroni adjustment *.

CTABLES

  /VLABELS VARIABLES=Gruppe Result DISPLAY=DEFAULT

  /TABLE Result [COUNT F40.0, COLPCT.COUNT PCT40.1] BY Gruppe

  /SLABELS POSITION=ROW

  /CATEGORIES VARIABLES=Gruppe Result ORDER=A KEY=VALUE EMPTY=INCLUDE

  /COMPARETEST TYPE=PROP ALPHA=0.05 ADJUST=BONFERRONI ORIGIN=COLUMN

  INCLUDEMRSETS=YES CATEGORIES=ALLVISIBLE.