Output line without information
liangyy opened this issue · 4 comments
Hi Natsuhiko,
Occasionally, I got output like the following:
1 rs 1 0 N N -1.0 -1.0 -1.0 0.0 -1.0 -1.0 -1.0 -1.0 -1.0 -1 2 1 -1 -1 -1 -1 -1 0.0
It seems that it ran through null and alternative model according to log file as follow (running on one test SNP with two feature SNP)
#################################
# RASQUAL v 1.0 #
#################################
Hyper-parameters :
ab_phi: 10.001000
ab_pi : 10.001000
kappa : 2.020000
omega : 0.200000
beta0 : 100000.000000
sigma2: 10000.000000
Sample Size : 200
Init No. fSNPs : 2
Init No. testint SNPs : 3
Kth feature : 1
100
200
Effective feature length : 10000
Memory has been allocated...
heter_var1 100 km=0.690664 af=0.312500 hwe=41.322314 asgaf=0.500000 asaf=0.498920
RSQ=1.000000 EP=0.000000
heter_var2 200 km=0.138730 af=0.060000 hwe=0.814848 asgaf=0.500000 asaf=0.493646
RSQ=1.000000 EP=0.000000
Data has been loaded...
No. fSNPs : 2
No. all SNPs : 3 3
No. testing SNPs: 1
No ase/ase ratio: 0.000000
va0=5270392.205963 mu0=4915.938200 th0=4.589604
N0=200 P=1 beta=8.500238 th=4.589604 s=10000.000000 kappa=2.020000 omega=0.200000
beta0: 8.500238
init beta: 8.500238,
Init likelihood: -1818.862952 gradb=0.000000
mean eta=8.500223 sd eta=-nan sdki=0.000507 mki=8.500223
convstatus=1 200.000000 stuck=0.000000Final likelihood: -1818.672115
Final theta: 4.874713
Final beta: 0 8.500223
Gradient th: 0.000000
Gradient be: 0.000000
Glm. results:
Lkhd GLM=-1818.672115 Theta=4.874713
8.500223
10011.000000 6268.000000 5228.000000 3378.000000 4848.000000 5189.000000 3109.000000 1797.000000 5188.000000 8745.000000 6099.000000 4713.000000 7091.000000 2145.000000 3142.000000 2216.000000 9349.000000 3249.000000 3847.000000 4452.000000 5076.000000 3476.000000 6405.000000 3852.000000 2297.000000 5627.000000 3275.000000 8271.000000 7235.000000 1294.000000 5247.000000 8161.000000 2705.000000 3953.000000 3942.000000 11778.000000 3933.000000 5353.000000 3769.000000 4687.000000 4709.000000 3020.000000 5400.000000 5575.000000 9009.000000 2114.000000 6091.000000 3217.000000 4772.000000 4711.000000 3368.000000 6272.000000 4484.000000 2819.000000 5823.000000 8490.000000 2885.000000 1626.000000 4039.000000 5681.000000 3014.000000 7146.000000 1501.000000 2398.000000 4408.000000 3926.000000 5643.000000 1557.000000 6734.000000 6684.000000 7690.000000 5468.000000 5364.000000 2433.000000 4430.000000 6716.000000 6386.000000 4930.000000 7985.000000 3847.000000 7014.000000 3062.000000 5613.000000 2767.000000
0.999713 0.999961 1.000873 0.999869 1.000015 0.999941 0.999536 0.998969 0.999522 0.999866 1.000039 0.999196 1.000505 0.999615 0.999243 1.000348 0.999589 0.999409 0.999669 0.999510 0.999800 0.999987 0.999513 0.999513 1.000216 1.000262 1.000264 1.000175 1.000510 1.000518 1.000250 0.999941 0.998979 0.999721 0.999212 0.999963 1.000535 1.000489 1.000356 0.999679 1.000227 1.000447 1.000436 1.000219 1.000168 0.999083 0.999921 0.999046 0.999778 0.999276 1.000727 0.999790 1.000329 1.000021 1.000269 1.000296 1.000017 1.000115 0.999637 1.000486 0.999774 1.000696 0.999758 0.999720 1.000804 1.000627 0.999569 1.000484 0.998979 0.999318 0.999253 0.999938 1.001283 1.000311 0.999999 1.000389 1.000409 1.000829 1.000671 0.999817 0.999706 1.000265 0.999421 0.999998
4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999 4915.864999
Grand Null w/o AS [1] pi=0.500000 delta=0.010000 phi=0.500000 beta=8.500223 theta=4.874713 theta2=4.874713 asr=0.000000 lkhd=-1843.684188 ldiff=0.000000 gpi=0.000000 gd=0.000000 gh=0.000000 gb=-0.000000 gt=0.000000 gR=0.000000 tval=0.000000 kld=0.000000
Grand Null with AS [4] pi=0.500000 delta=0.010066 phi=0.503711 beta=8.500223 theta=7.358195 theta2=7.358195 asr=0.000000 lkhd=-2355.031560 ldiff=0.000000 gpi=0.000000 gd=0.000000 gh=-0.000000 gb=-0.000000 gt=-0.000000 gR=0.000000 tval=0.000000 kld=0.000000
Grand Null with AS [231] pi=0.500000 delta=0.000000 phi=0.503971 beta=8.500223 theta=7.373650 theta2=7.373650 asr=0.000000 lkhd=-2371.397468 ldiff=0.055857 gpi=0.000000 gd=-0.990000 gh=-12.903892 gb=0.000000 gt=-10.653832 gR=0.000000 tval=0.000000 kld=0.000000
Grand Null with AS [231] pi=0.500000 delta=0.000000 phi=0.478140 beta=8.500223 theta=9.494437 theta2=9.494437 asr=0.000000 lkhd=-2364.384132 ldiff=0.065054 gpi=0.000000 gd=0.010000 gh=60.307564 gb=0.000000 gt=-52.711117 gR=0.000000 tval=0.000000 kld=0.000000
Grand Null with AS [12] pi=0.500000 delta=0.010068 phi=0.503711 beta=8.500223 theta=7.358195 theta2=7.358195 asr=0.000000 lkhd=-2355.031560 ldiff=0.000000 gpi=0.000000 gd=-0.000001 gh=0.000000 gb=-0.000000 gt=0.000002 gR=0.000000 tval=0.000000 kld=0.000000
Grand Null with AS [17] pi=0.500000 delta=0.010071 phi=0.503711 beta=8.500223 theta=7.358195 theta2=7.358195 asr=0.000000 lkhd=-2355.031560 ldiff=0.000002 gpi=0.000000 gd=-0.000004 gh=0.000000 gb=-0.000000 gt=0.000006 gR=0.000000 tval=0.000000 kld=0.000000
Grand Null with AS [15] pi=0.500000 delta=0.010067 phi=0.503711 beta=8.500223 theta=7.358195 theta2=7.358195 asr=0.000000 lkhd=-2355.031560 ldiff=0.000000 gpi=0.000000 gd=-0.000001 gh=0.000000 gb=-0.000000 gt=0.000002 gR=0.000000 tval=0.000000 kld=0.000000
Fitting null hypothesis finished.
Alternative Start
2 testSNP 4.874713 [4] pi=0.537175 delta=0.010000 phi=0.500000 beta=8.555522 theta=5.194722 theta2=5.194722 asr=0.000000 lkhd=-1842.902025 ldiff=0.000000 gpi=-0.000004 gd=0.000000 gh=0.000000 gb=0.000000 gt=-0.000001 gR=0.000000 tval=0.000094 kld=0.982743
2 testSNP [231] pi=0.493666 delta=0.000000 phi=0.505819 beta=8.490609 theta=7.206939 theta2=7.206939 asr=0.000000 lkhd=-2355.800577 ldiff=0.063030 gpi=29.188328 gd=0.010000 gh=-4.318574 gb=0.699464 gt=1.396312 gR=0.000000 tval=0.000001 kld=0.982709
ASEQTLALL finished.
Could you clarify what happens here?
Thanks!
Yanyu
Hi Yanyu,
For some reason the EM algorithm didn't converge and you got the alternative likelihood (-2355.800577) lower than the null likelihood (-2355.031560), which should be greater all the time. In this case, RASQUAL does not report any valid result.
Best regards,
Natsuhiko
Hi Natsuhiko,
Thanks for your explanation. I have a follow-up question. Does it mean that somehow the alternative model gets stuck at a local optimum?
Thanks!
Yanyu
Hi Yanyu,
It did not even converge to a local optimum. It reached the maximum iterations and the EM algorithm couldn't find any solution.
Best regards,
Natsuhiko
Hi Natsuhiko,
Thanks for the explanation! I'll close the issue.
Yanyu