Errore in più misure ripetute

Question

Sto provando a eseguire un modello misto lineare con misure ripetute a 57 diversi punti di tempo. Ma continuo a ricevere il messaggio di errore:

Error in solve.default(estimates[dimE[1L] - (p:1), dimE[2L] - (p:1), drop = FALSE]) : 

sistema è computazionalmente singolare: reciproca numero condition = 7.7782e-18

Cosa significa?

Il mio codice è tale:

model.dataset = data.frame(TimepointM=timepoint,SubjectM=sample,GeneM=gene) 
library("nlme") 
model = lme(score ~ TimepointM + GeneM,data=model.dataset,random = ~1|SubjectM)  

Ecco i dati:

score = c(2,-3,11,14,1,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,6,7,2,-3,11,14,1,7,6,7,2,-3,11,14,1,7,6,2,-3,11,14,1,7,7,2,-3,11,14,1,7,6,2,-3,11,14,1,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,2,-3,11,14,1,7,6,2,-3,11,14,1,7,7,2,-3,11,14,1,7,6,7,2,-3,11,14,1,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,6,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,6,7,2,-3,11,14,1,7,6,2,-3,11,14,1,7,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7,2,-3,11,14,1,7) 
timepoint = c(1,1,1,1,1,1,2,2,2,2,2,2,3,3,3,3,3,3,3,4,4,4,4,4,4,5,5,5,5,5,5,6,6,6,6,6,6,7,7,7,7,7,7,8,8,8,8,8,8,8,9,9,9,9,9,9,10,10,10,10,10,10,10,12,12,12,12,12,12,12,12,13,13,13,13,13,13,13,13,14,14,14,14,14,14,14,14,15,15,15,15,15,15,16,16,16,16,16,16,16,16,17,17,17,17,17,17,18,18,18,18,18,18,19,19,19,19,19,19,19,20,20,20,20,20,20,21,21,21,21,21,21,24,24,24,24,24,24,24,25,25,25,25,25,25,25,27,27,27,27,27,27,28,28,28,28,28,28,29,29,29,29,29,29,30,30,30,30,30,30,30,31,31,31,31,31,31,31,32,32,32,32,32,32,33,33,33,33,33,33,33,33,34,34,34,34,34,34,34,35,35,35,35,35,35,36,36,36,36,36,36,36,37,37,37,37,37,37,38,38,38,38,38,38,39,39,39,39,39,39,39,40,40,40,40,40,40,40,41,41,41,41,41,41,41,41,42,42,42,42,42,42,42,42,43,43,43,43,43,43,44,44,44,44,44,44,44,45,45,45,45,45,45,46,46,46,46,46,46,47,47,47,47,47,47,48,48,48,48,48,48,49,49,49,49,49,49,49,50,50,50,50,50,50,51,51,51,51,51,51,52,52,52,52,52,52,52,53,53,53,53,53,53,53,54,54,54,54,54,54,55,55,55,55,55,55,56,56,56,56,56,56,57,57,57,57,57,57) 
sample = c("S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S13T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S13T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S13T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S13T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S13T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S13T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S13T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S13T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S13T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S01T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0","S02T0","S03T0","S07T0","S09T0","S10T0","S12T0") 
gene =c(24.1215870,-18.8771658,-27.3747309,-41.5740199,26.1561877,-2.7836332,20.8322796,36.5745088,-24.1541743,-11.2362216,4.9042852,7.4230219,155.8663563,16.4465366,-11.7982286,-1.6102783,-35.9559091,27.7909495,-13.9181661,-29.6037658,-68.4297261,-45.0877920,-48.3157529,17.1649982,-26.9084544,19.7358439,-5.8991143,-24.1541743,-23.5960654,13.0780939,-2.7836332,18.6394081,-28.3157487,-49.9186269,-33.7086648,41.6864242,-30.6199654,36.1823804,-36.5745088,-49.9186269,-44.9448864,-4.9042852,-34.3314764,62.3465425,-42.7609951,-11.7982286,-32.2055657,-56.1811080,5.7216661,-17.6296771,4.3857431,-43.6534459,9.6616697,-44.9448864,18.7997599,-12.9902884,109.1064494,7.6750504,-43.6534459,-17.7130611,-25.8433097,5.7216661,-18.5575548,35.2750175,36.1823804,2.3596457,-25.7644526,-55.0574858,15.5302365,-19.4854325,73.3687689,63.1668918,20.8322796,16.5175201,-22.5438960,-28.0905540,15.5302365,7.4230219,39.5062602,107.4657509,36.1823804,-23.5964573,-45.0877920,-43.8212642,4.0869043,-40.8266205,26.3375068,13.1572292,-25.9561030,-40.2569571,-52.8102415,2.4521426,-49.1775202,246.1047731,36.1823804,11.7982286,-35.4261223,-26.9669318,-2.4521426,-38.0429873,38.5656349,9.8679219,16.5175201,8.0513914,-42.6976421,26.9735686,-26.9084544,4.3857431,12.9780515,-32.2055657,-33.7086648,9.8085704,-36.2800196,215.7518511,6.5786146,-9.4385829,-19.3233394,-40.4503978,17.1649982,-7.4230219,14.2536650,-23.5964573,-53.1391834,-52.8102415,22.0692834,-54.7447866,24.1215870,-44.8332688,-24.1541743,-42.6976421,26.9735686,-40.8266205,191.1413737,17.5429723,-70.7893718,-37.0364006,-39.3267756,-4.9042852,-0.9278777,93.5198138,-6.5786146,-24.7762801,-28.9850091,-39.3267756,22.0692834,-50.1053979,14.2536650,23.5964573,-20.9336177,-53.9338637,14.7128556,-39.8987428,4.3857431,-64.8902575,-59.5802966,-33.7086648,22.0692834,2.7836332,46.0503024,-35.3946859,-43.4775137,-53.9338637,30.2430921,-34.3314764,80.3942259,28.5073300,-87.3068919,-24.1541743,-62.9228410,13.0780939,-25.0526990,35.0859447,-24.7762801,-38.6466789,-58.4283523,31.0604729,0.0000000,24.4562563,1.0964358,-27.1359259,-75.6830794,-16.8543324,20.4345217,-11.1345329,74.1390629,18.2282447,-27.3044720,-45.2890768,-46.7707724,15.3258912,-27.9523169,-6.9763039,117.3099418,18.6394081,-21.2368115,-38.6466789,-34.8322870,22.0692834,-48.2496425,6.5786146,-64.8902575,-51.5289052,-80.9007955,23.7040451,-26.9084544,223.1349942,8.7714862,10.6184058,-127.2119846,-31.4614205,0.8173809,-16.7017993,9.8679219,-35.3946859,-54.7494617,-44.9448864,14.7128556,-18.5575548,97.5827836,-166.3550237,-95.0064189,-123.5984376,104.6247509,-121.5519839,33.9895089,-44.8332688,-40.2569571,-56.1811080,51.4949946,0.0000000,-16.9312544,95.9808615,6.5786146,-21.2368115,-9.6616697,-13.4834659,10.6259513,-25.9805767,116.4895926,-1.0964358,-16.5175201,-56.3597400,-44.9448864,13.8954747,-12.9902884,-5.6437515,71.3703842,25.2180227,-41.2938002,-53.1391834,-32.5850426,8.9911895,12.9902884,31.9812582,1.0964358,-70.7893718,-33.8158440,-38.2031534,-15.5302365,-25.0526990,153.4053085,36.1823804,-34.2148630,-41.8672354,-19.1015767,22.8866643,0.9278777,20.8322796,-29.4955716,-43.4775137,-69.6645739,33.5126155,-45.4660092,26.3144585,-33.0350402,24.1541743,-42.6976421,0.0000000,-28.7642099,38.3752520,-7.0789372,-22.5438960,-20.2251989,34.3299964,19.4854325,4.3857431,-61.3507889,-33.8158440,-64.0464631,39.2342816,-28.7642099,183.7582306,-4.3857431,-22.4166344,-28.9850091,-57.3047302,25.3388069,-26.9084544,35.0859447,7.0789372,-33.8158440,-43.8212642,-1.6347617,5.5672664,-35.0859447,-40.1139773,-14.4925046,-12.3598438,21.2519025,-14.8460438,119.7709896,30.7002016,-22.4166344,-46.6980703,-43.8212642,5.7216661,-10.2066551,203.4466124,116.2221917,-83.7674233,-109.4989234,-38.2031534,78.4685632,-56.6005421,21.9287154,-63.7104346,-56.3597400,-4.4944886,25.3388069,-73.3023414,29.6037658,-31.8552173,-46.6980703,-79.7771734,21.2519025,-18.5575548,16.4465366,-27.1359259,-43.4775137,-41.5740199,-11.4433321,-23.1969435,27.4108943,-84.9472461,-53.1391834,-40.4503978,22.8866643,16.7017993) 

Answer 1

tl; dr Credo che il problema è che ogni individuo ha esattamente lo stesso valore di risposta (score) per ogni punto temporale (ovvero perfetta omogeneità all'interno degli individui), quindi il termine degli effetti casuali spiega completamente i dati; non è rimasto nulla per gli effetti fissi. Sei sicuro di non voler usare gene come variabile di risposta ?? (Scoperto dopo l'esecuzione attraverso un mazzo di modellazione tentativi, da tracciando i dati maledetti, qualcosa che tutti dovrebbero fare sempre prima ...)

## simplifying names etc. slightly 
dd <- data.frame(timepoint,sample,gene,score,) 
library("nlme") 
m0 <- lme(score ~ timepoint + gene, data=dd, 
    random = ~1|sample) 
## reproduces error 

Come un primo controllo, facciamo solo vedere se c'è qualcosa nella vostra fissa modello di -Effetto che è singolare:

lm(score~timepoint+gene,dd) 
## 
## Call: 
## lm(formula = score ~ timepoint + gene, data = dd) 
## 
## Coefficients: 
## (Intercept) timepoint   gene 
## 5.414652 -0.004064 -0.024485 

No, che funziona bene.

Proviamo a lme4:

library(lme4) 
m1 <- lmer(score ~ timepoint + gene + (1|sample), data=dd) 
## Error in fn(x, ...) : Downdated VtV is not positive definite 

Proviamo ridimensionamento & centrare i dati - a volte questo aiuta:

ddsc <- transform(dd, 
    timepoint=scale(timepoint), 
    gene=scale(gene)) 

lme fallisce ancora:

m0sc <- lme(score ~ timepoint + gene, data=ddsc, 
    random = ~1|sample) 

lmer opere -- S o di!

m1sc <- lmer(score ~ timepoint + gene + (1|sample), data=ddsc) 
## Warning message: 
## In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : 
## Model is nearly unidentifiable: very large eigenvalue 
## - Rescale variables? 

I risultati forniscono coefficienti per i parametri che sono quasi azzerati. (La varianza residua è anche infinitamente piccolo.)

## m1sc 
## Linear mixed model fit by REML ['lmerMod'] 
## Formula: score ~ timepoint + gene + (1 | sample) 
## Data: ddsc 
## REML criterion at convergence: -9062.721 
## Random effects: 
## Groups Name  Std.Dev. 
## sample (Intercept) 7.838e-01 
## Residual    3.344e-07 
## Number of obs: 348, groups: sample, 8 
## Fixed Effects: 
## (Intercept) timepoint   gene 
## 5.714e+00 -4.194e-16 -1.032e-14 

A questo punto posso solo pensare a un paio di possibilità:

• c'è qualcosa circa la progettazione che significa che gli effetti casuali sono in qualche modo (?) completamente confuso con uno o entrambi gli effetti fissi
• questi sono dati simulati che sono costruiti artificialmente per essere perfettamente bilanciati ...?

library(ggplot2); theme_set(theme_bw()) 
ggplot(dd,aes(timepoint,score,group=sample,colour=gene))+ 
     geom_point(size=4)+ 
     geom_line(colour="red",alpha=0.5) 

Aha!

Answer 2

In modo che R risolva una matrice, deve essere calcolabile in modo invertibile. L'errore che stai recuperando ti sta dicendo che, per motivi computazionali, la tua matrice è singolare, il che significa che non ha un inverso.

Poiché questo errore riguarda più il lato della teoria statistica, è probabilmente più adatto per la convalida incrociata. Vedere this link per ulteriori informazioni.

Controlla i tuoi dati per assicurarti di non avere variabili indipendenti perfettamente correlate.