I am conducting the following LMEM with a brain volume (continuous numerical) as my DV, Group (categorical Control versus Autistic), Hemisphere (categorical Left versus Right), SEX (categorical Male versus Female) and Age (continuous numerical) as my IVs. I want to include MRI scanner site (categorical, 15 different sites) as my random effect. The following model works well:
Putamen_model <- lmer(Putamen ~ DX_GROUP * Hemisphere * AGE_AT_SCAN * SEX
+ (1|SITE_ID),
data = L_Putamen_Outlier_DF)
However, i want to take into account the fact that each participant has two putamen values, one for the right hemisphere and one for the left hemisphere in my model. I am not sure if this is the right thing to do, but I believe I need to indicate that hemisphere is nested in my model with the following model:
Putamen_model_Age <- lmer(Putamen ~ DX_GROUP * Hemisphere * AGE_AT_SCAN * SEX + (1|Hemisphere/SUB_ID)
+ (1 |SITE_ID2),
data = L_Putamen_Outlier_DF)
Yet I get the following error because I have as many hemisphere values as participants:
Error: number of levels of each grouping factor must be < number of observations
> table(L_Putamen_Outlier_DF$Hemisphere)
Left Right
644 644
> summary(table(L_Putamen_Outlier_DF$SUB_ID))
Number of cases in table: 1288
Number of factors: 1
> summary(L_Putamen_Outlier_DF$SUB_ID)
50002 50004 50005 50006 50012 50022 50023 50024 50030 50031 50032 50045 50046 50047 50048 50049 50051 50053 50054
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
50056 50059 50102 50103 50104 50105 50106 50109 50110 50111 50112 50113 50114 50116 50120 50121 50122 50124 50127
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
50128 50129 50130 50131 50134 50137 50142 50144 50145 50147 50149 50150 50152 50153 50157 50158 50160 50161 50162
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
50163 50164 50165 50166 50167 50168 50170 50171 50186 50187 50188 50189 50190 50193 50194 50195 50196 50198 50199
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
50200 50201 50202 50203 50204 50205 50206 50208 50209 50210 50211 50212 50213 50214 50215 50216 50232 50241 50245
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
50250 50251 50254 50255 (Other)
2 2 2 2 1090
I clearly see that there seems to be as many subjects as observations in my case and that Linear mixed effect model supposes that there is less subjects than observations so it throws an error if it is not the case. However I want to indicate that for one participant I have 2 volume values, one for each hemisphere.
How can I tell the model that I have 2 values for each participant, one for each hemisphere? Can you suggest any models? Any help is welcome.
I also tried including (1 | SUB_ID) + (1 | SITE_ID2) but my model fails to converge.
Thank you, Camille
Aucun commentaire:
Enregistrer un commentaire