Dear Neil
Thanks for the advice. Here is the output for other readers. "Rubbish" refers to the OR for variable cox2use and
c._hap11*c.cox2use...
. xi: haplologit stroke i.cox2use i.aspirin i.nsaid i.gender i.agegroup2 i.bmigroup2 i.Diabetes4 i.HT
> i.AF2 ///
> (i.Smoker) if ~(inlist(kozak,.) & inlist(thrmet,.)), snpvars(kozak thrmet) ///
> riskhap1("11", inter(cox2use)) ///
> inher(d) happrefix(_hap) or
i.cox2use _Icox2use_0-1 (naturally coded; _Icox2use_0 omitted)
i.aspirin _Iaspirin_0-1 (naturally coded; _Iaspirin_0 omitted)
i.nsaid _Insaid_0-1 (naturally coded; _Insaid_0 omitted)
i.gender _Igender_1-2 (_Igender_1 for gender==F omitted)
i.agegroup2 _Iagegroup2_1-2 (naturally coded; _Iagegroup2_1 omitted)
i.bmigroup2 _Ibmigroup2_1-2 (naturally coded; _Ibmigroup2_1 omitted)
i.Diabetes4 _IDiabetes4_0-1 (naturally coded; _IDiabetes4_0 omitted)
i.HT _IHT_0-1 (naturally coded; _IHT_0 omitted)
i.AF2 _IAF2_0-1 (naturally coded; _IAF2_0 omitted)
i.Smoker _ISmoker_1-3 (naturally coded; _ISmoker_3 omitted)
Building consistent haplotype pairs:
Obtaining initial haplotype frequency estimates from the control sample:
Haplotype frequency EM estimation
Number of iterations = 18
Sample log-likelihood = -479.73979
+------------------------+
| haplotype | frequency* |
|-----------+------------|
| 00 | .003216 |
| 01 | .11588 |
| 10 | .062492 |
| 11 | .818411 |
+------------------------+
* frequencies > .0020263
Performing gradient-based optimization:
Iteration 0: Retrosp. profile log likelihood = -1643.8538
Iteration 1: Retrosp. profile log likelihood = -1453.0153
Iteration 2: Retrosp. profile log likelihood = -1451.7071
Iteration 3: Retrosp. profile log likelihood = -1451.6272
Iteration 4: Retrosp. profile log likelihood = -1451.6222
Iteration 5: Retrosp. profile log likelihood = -1451.6211
Iteration 6: Retrosp. profile log likelihood = -1451.6208
Iteration 7: Retrosp. profile log likelihood = -1451.6207
Iteration 8: Retrosp. profile log likelihood = -1451.6207
Haplotype-effects logistic regression
Mode of inheritance: dominant Number of obs = 987
Genetic distribution: Hardy-Weinberg equilib. Number phased = 973
Genotype: kozak thrmet Number unphased = 14
Number missing = 0
Wald chi2(13) = 247.92
Retrosp. profile log likelihood = -1451.6207 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
stroke | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
_Icox2use_1 | 3.05e-06 .0017547 -0.02 0.982 0 .
_Iaspirin_1 | 1.787856 .3324477 3.12 0.002 1.241806 2.574018
_Insaid_1 | 1.017781 .253331 0.07 0.944 .6248646 1.657765
_Igender_2 | 1.446311 .2393843 2.23 0.026 1.045626 2.00054
_Iagegroup~2 | 9.470277 1.765575 12.06 0.000 6.571589 13.64756
_Ibmigroup~2 | .5589643 .0981841 -3.31 0.001 .3961572 .7886797
_IDiabetes~1 | 1.853561 .3687461 3.10 0.002 1.255074 2.737438
_IHT_1 | 2.43637 .4227256 5.13 0.000 1.734021 3.423198
_IAF2_1 | 2.12824 .4636764 3.47 0.001 1.388579 3.261899
_ISmoker_1 | 6.209872 1.606923 7.06 0.000 3.739538 10.3121
_ISmoker_2 | 1.074069 .1920307 0.40 0.689 .756567 1.524814
_hap11 | 1.03895 .3069367 0.13 0.897 .5822703 1.853809
|
c._hap11* |
c.cox2use | 170300.5 9.80e+07 0.02 0.983 0 .
------------------------------------------------------------------------------
Haplotype frequencies
------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
_hap00 | .0016101 .0014436 1.12 0.265 -.0012194 .0044395
_hap01 | .1129663 .007536 14.99 0.000 .0981959 .1277366
_hap10 | .0650989 .0057849 11.25 0.000 .0537607 .0764371
_hap11 | .8203248 .0094228 87.06 0.000 .8018565 .838793
------------------------------------------------------------------------------
>>> Neil Shephard <[email protected]> 26/08/2009 5:47 pm >>>
On Wed, Aug 26, 2009 at 5:39 AM, Jane
Maguire<[email protected]> wrote:
> Dear all
> Data is case control, snps are coded 0,1,2
> One gene, two snps. Haplotype for hap00 is quite rare and when interaction term (cox2use) is added the output is rubbish.
What does "rubbish" mean?
> I am not a biostatistician- am a Phd student in medicine and Public Health.
You'd benefit from getting your head round the basics of statistics in
the long run. A good book is Intuitive Biostatistics by Harvey
Motulsky. Not too heavy on the maths, but explains things very well.
> However, I think the problem is the numbers are too small...unless I have made a coding error? any advice?
>
> This command works- without interaction,
>
> xi: haplologit stroke i.cox2use i.aspirin i.nsaid i.gender i.agegroup2 i.bmigroup2 i.Diabetes4 i.HT i.AF2 ///
> (i.Smoker) if ~(inlist(kozak,.) & inlist(thrmet,.)), snpvars(kozak thrmet) inher(d) ///
> riskhap1("11") ///
> happrefix(_hap) or noemtable noemshow
>
> But this one doesn't,
>
> set more off
> xi: haplologit stroke i.cox2use i.aspirin i.nsaid i.gender i.agegroup2 i.bmigroup2 i.Diabetes4 i.HT i.AF2 ///
> (i.Smoker) if ~(inlist(kozak,.) & inlist(thrmet,.)), snpvars(kozak thrmet) ///
> riskhap1("11", inter(cox2use)) ///
> inher(d) happrefix(_hap) or
As per above, what does "rubbish" mean? If you could show the output
of running these two commands it would be more informative for those
on the list who might then be able to comment on what the problem is.
Neil
--
"The combination of some data and an aching desire for an answer does
not ensure that a reasonable answer can be extracted from a given body
of data." ~ John Tukey (1986), "Sunset salvo". The American
Statistician 40(1).
Email - [email protected]
Website - http://slack.ser.man.ac.uk/
Photos - http://www.flickr.com/photos/slackline/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
