turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- Wald Test results inconsistent with test of GEE pa...

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

02-22-2010 01:27 PM

Hi there,

I ran genmod on a binary DV, and I was puzzled that the parameter for my "time2" variable was nonsignificant, whereas the Wald test for this same variable was significant. I show my syntax and output below. Could anyone tell me why these results differed (the other parameters matched the p values of the Wald test)? For some background, I am working with a study that consisted of 2 phases in time. The first went from 0 to 16 weeks, and the second phase went from 16 to 64 weeks. That is why I form a vector representing these time points below, because I needed to have 2 separate slopes to model the differences between these 2 differing phases.

data ham;

set ham;

*set up vectors of time points;

if time le 16 then time1 = time;

else time1 = 16;

if time gt 16 then time2 = time;

else time2 = 0;

run;

proc genmod data=ham descending;

class id hasAnxB time;

model MDD_C = time1 time2 hasAnxB time1*hasAnxB time2*hasAnxB /link=logit dist=binomial type3 wald;

repeated subject=id / within=time corrw type=un;

run;

The GENMOD Procedure

Analysis Of GEE Parameter Estimates

Empirical Standard Error Estimates

Parameter Estimate S.E. Z Pr > |Z|

Intercept 1.6477 0.3973 0.8691 2.4263 4.15 <.0001

time1 -0.1792 0.0302 -0.2383 -0.1201 -5.94 <.0001

time2 -0.0021 0.0049 -0.0117 0.0074 -0.44 0.6623

HASANXB 0.00 -0.7935 0.4918 -1.7574 0.1703 -1.61 0.1066

HASANXB 1.00 0.0000 0.0000 0.0000 0.0000 . .

time1*HASANXB 0.00 0.0724 0.0392 -0.0043 0.1492 1.85 0.0644

time1*HASANXB 1.00 0.0000 0.0000 0.0000 0.0000 . .

time2*HASANXB 0.00 -0.0269 0.0107 -0.0478 -0.0059 -2.52 0.0119

time2*HASANXB 1.00 0.0000 0.0000 0.0000 0.0000 . .

Wald Statistics For Type 3 GEE Analysis

Source DF Square Pr > ChiSq

time1 1 53.31 <.0001

time2 1 8.50 0.0036

HASANXB 1 2.60 0.1066

time1*HASANXB 1 3.42 0.0644

time2*HASANXB 1 6.33 0.0119

Message was edited by: mburns

Message was edited by: mburns

Message was edited by: mburns

Message was edited by: mburns Message was edited by: mburns

I ran genmod on a binary DV, and I was puzzled that the parameter for my "time2" variable was nonsignificant, whereas the Wald test for this same variable was significant. I show my syntax and output below. Could anyone tell me why these results differed (the other parameters matched the p values of the Wald test)? For some background, I am working with a study that consisted of 2 phases in time. The first went from 0 to 16 weeks, and the second phase went from 16 to 64 weeks. That is why I form a vector representing these time points below, because I needed to have 2 separate slopes to model the differences between these 2 differing phases.

data ham;

set ham;

*set up vectors of time points;

if time le 16 then time1 = time;

else time1 = 16;

if time gt 16 then time2 = time;

else time2 = 0;

run;

proc genmod data=ham descending;

class id hasAnxB time;

model MDD_C = time1 time2 hasAnxB time1*hasAnxB time2*hasAnxB /link=logit dist=binomial type3 wald;

repeated subject=id / within=time corrw type=un;

run;

The GENMOD Procedure

Analysis Of GEE Parameter Estimates

Empirical Standard Error Estimates

Parameter Estimate S.E. Z Pr > |Z|

Intercept 1.6477 0.3973 0.8691 2.4263 4.15 <.0001

time1 -0.1792 0.0302 -0.2383 -0.1201 -5.94 <.0001

time2 -0.0021 0.0049 -0.0117 0.0074 -0.44 0.6623

HASANXB 0.00 -0.7935 0.4918 -1.7574 0.1703 -1.61 0.1066

HASANXB 1.00 0.0000 0.0000 0.0000 0.0000 . .

time1*HASANXB 0.00 0.0724 0.0392 -0.0043 0.1492 1.85 0.0644

time1*HASANXB 1.00 0.0000 0.0000 0.0000 0.0000 . .

time2*HASANXB 0.00 -0.0269 0.0107 -0.0478 -0.0059 -2.52 0.0119

time2*HASANXB 1.00 0.0000 0.0000 0.0000 0.0000 . .

Wald Statistics For Type 3 GEE Analysis

Source DF Square Pr > ChiSq

time1 1 53.31 <.0001

time2 1 8.50 0.0036

HASANXB 1 2.60 0.1066

time1*HASANXB 1 3.42 0.0644

time2*HASANXB 1 6.33 0.0119

Message was edited by: mburns

Message was edited by: mburns

Message was edited by: mburns

Message was edited by: mburns Message was edited by: mburns

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

02-24-2010 12:23 PM

The test in the "Analysis of GEE Parameter Estimates" table is a test that the individual parameter is zero, while the test in the "Type 3 GEE analysis" table is a test of a type3 hypothesis. A type3 hypothesis for a model effect might involve other parameters in the model, such as when it is involved in interactions.

To see the form of the type3 hypothesis for a given effect, run the same model in PROC GLM (omit the REPEATED statement) and specify the E and E3 options in the MODEL statement of PROC GLM (ignore all GLM results other than the Estimable Functions tables). The E option shows the general form of the estimable functions, and the E3 option shows the form of the type3 hypothesis that is tested by the TYPE3 option in PROC GENMOD and displayed in the Type 3 analysis table.

To see the form of the type3 hypothesis for a given effect, run the same model in PROC GLM (omit the REPEATED statement) and specify the E and E3 options in the MODEL statement of PROC GLM (ignore all GLM results other than the Estimable Functions tables). The E option shows the general form of the estimable functions, and the E3 option shows the form of the type3 hypothesis that is tested by the TYPE3 option in PROC GENMOD and displayed in the Type 3 analysis table.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

02-24-2010 06:51 PM

Great to know, thank you!