Category Error when using FitGLME

So I have several data sets with ordtinal and categorical values, I see no big difference in them. However, some of hese data sets give no errors while others give me this strange value for which I can find no explanation. Anyone have this issue? By the way, taking out that categorical variable didn't solve the issue.

Thanks!

Error using categorical (line 434)

Unable to create default category names. Specify category names using the CATEGORYNAMES input argument.

Error in nominal (line 152)

b = b@categorical(a,args{:},'Ordinal',false);

Error in classreg.regr.LinearLikeMixedModel/makeInteractionVar (line 806)

G = nominal(ds.(interactionVars{1}));

Error in classreg.regr.LinearLikeMixedModel/extractGroupingInfo (line 205)

model.makeInteractionVar(ds,interactionVars);

Error in GeneralizedLinearMixedModel.fit (line 2377)

model.GroupingInfo = extractGroupingInfo(model,dssubset);

Error in fitglme (line 398)

glme = GeneralizedLinearMixedModel.fit(T,formula,varargin{:});

Error in GLMFinal (line 35)

glmeVw2 = fitglme(bat1,'Vw ~ Td + (1|ThetaX) + (1|ThetaY) + (1|ThetaZ)','FitMethod','Laplace')

[Ive J]

Please share your script along with the dataset you use (or at least a simulated one or more info on your data structure).

[Walter Roberson]

Do the variable names Vw, Td, ThetaX, ThetaY, and ThetaZ all exist in your table bat1 ?

[Sam Litvin]

Ive Ive

function GLMFinal(bat,noL)
%GLM of combined files for each bats Thetas and Trident distance versus
%Vertical and horizontal beam widths and beam direction
%If you do not want to use data for L, enter 1, otherwise 0 for nol
%input is taken from GLMFinal file
%output saved as bat_pvals in the pvals folder
 
dirf='Results/';
 
if bat==2
        bat='2';
    elseif bat==5
        bat='5';
    elseif bat==6
        bat='6';
    elseif bat==9
        bat='9'; 
end
 
bat1=load([dirf,'GLM_',bat,'.mat']);
bat1=struct2table(bat1.b4)
bat1.Td;
bat1.Ph=categorical(bat1.Ph);
 
%bat1.Ph = categorical(bat1.Ph);
%bat1.Ph;
% 
%  lmeVw = fitlme(bat1,'Vw ~ Td + (1|Ph)+ (1|ThetaX) + (1|ThetaY) + (1|ThetaZ)', 'DummyVarCoding', 'reference'); 
%  glmeVd = fitlme(bat1,'Vd ~ 1 + Td + (ThetaX) + (ThetaY) + (ThetaZ) + (Ph)')
%  glmeHw = fitlm(bat1,'Hw ~ 1 + Td + (ThetaX) + (ThetaY) + (ThetaZ) + (Ph)')
% glmeHd = fitlm(bat1,'Hd ~ 1 + Td + (ThetaX) + (ThetaY) + (ThetaZ) + (Ph)')
% anova(lmeVw)
 
%Fitting the model and getting p values for angles and trident.
glmeVw2 = fitglme(bat1,'Vw ~ Td + (1|ThetaX) + (1|ThetaY) + (1|ThetaZ)','FitMethod','Laplace')
    glmeVw = fitglme(bat1,'Vw ~ Td + (1|Ph)+ (1|ThetaX) + (1|ThetaY) + (1|ThetaZ)+ (1+Td|ThetaX)+(1+Td|ThetaY)','FitMethod','Laplace')

Answer 1

[Sam Litvin]

yes, I print the table, prior to using fitglm to make sure, like i say, it works on one data set but not on another, the data sets look the same.

this set doesn't work:

ThetaX ThetaY ThetaZ Td Vd Vw Hw Hd Ld Lw

________ ________ _______ _______ _______ ____ ____ _______ _______ ____

1.2333 1.2445 0.52498 0.53232 6.5127 34.5 71.5 -15.891 NaN NaN

1.6859 2.1313 0.93286 0.90519 6.5127 37.5 NaN NaN NaN NaN

1.3092 0.5145 3.7508 0.58472 29.102 41 55 -15.891 NaN NaN

-0.52485 1.4835 4.1045 1 29.102 37 NaN NaN NaN NaN

-11.536 -4.3153 13.498 0 36.438 38 NaN NaN NaN NaN

-4.6048 -3.5271 7.2064 0.55741 14.538 41.5 NaN NaN NaN NaN

13.477 -2.3554 2.5966 0 -1.105 39 35 15.911 15.182 14.5

26.936 -4.7887 5.1617 0.84217 -18.806 40 39 15.911 15.182 37.5

26.93 -4.7895 5.1617 0.84233 -1.105 42 45.5 15.911 15.182 45.5

26.926 -4.7841 5.1623 0.84255 -1.105 33.5 28 15.911 15.182 41

26.922 -4.7668 5.1672 0.84266 -1.105 34 46.5 31.931 -33.431 27

27.116 -4.4235 5.3232 0.85596 6.448 29 44.5 15.911 -23.12 47

27.539 -4.2149 5.3918 0.85807 6.448 33 52.5 15.911 NaN NaN

27.843 -4.0407 5.5783 0.83484 -1.105 35 50.5 15.911 -23.12 41.5

28.616 -3.5122 6.3644 0.76825 -1.105 31.5 46.5 -16.014 NaN NaN

28.765 -3.3351 6.5384 0.77406 -1.105 39.5 43.5 -16.014 -23.12 46.5

30.946 -2.51 8.2508 0.64673 6.448 45.5 57 -9.2986 -33.431 35.5

37.91 1.9361 16.855 0.44011 -1.105 31 40 -16.014 15.182 60.5

42.227 3.8641 22.334 0.41069 -1.105 35 53 15.911 15.182 15.5

36.45 2.6909 16.286 0.43821 -1.105 35 58.5 15.911 15.182 44

31.617 -0.96164 9.4995 0.64966 -1.105 33.5 40.5 40.184 22.574 45.5

29.221 -3.4149 6.1208 0.86041 -1.105 32 38 31.931 31.814 32

This one does

bat1 =

87×11 table

ThetaX ThetaY ThetaZ Td Vd Vw Hw Hd Ld Lw Ph

_________ _________ _________ ________ _______ ____ _____ _______ _______ ____ __

1.1022 -1.0173 0.067847 0.60151 31.072 45 60.5 -16.492 NaN NaN 0

1.2061 -1.0606 0.16396 1 15.428 44.5 62.5 -16.492 NaN NaN 0

0.97804 -1.0005 0.19531 0.27491 6.9766 39.5 36.5 -16.492 NaN NaN 0

0.76398 -0.97725 0.1889 0 6.9766 41.5 12.5 -16.492 NaN NaN 0

0.0093436 0.27112 0.17762 0.099879 6.9766 29.5 34.5 -16.492 24.472 59.5 0

-0.059496 0.4879 0.39032 0.21028 22.845 34 85.5 -16.492 NaN NaN 0

0.12878 0.50104 0.34031 0.32137 31.072 12 14 -16.492 NaN NaN 0

-0.040724 0.63091 0.40665 0.20183 6.9766 23.5 35.5 -16.492 NaN NaN 0

0.17804 0.53765 0.25413 0.16974 22.845 19 31 -23.81 NaN NaN 0

-0.15602 0.4878 0.26046 0.16388 -1.2509 13.5 35.5 -16.492 NaN NaN 0

-0.25819 0.35799 0.31463 0.042781 6.9766 22 15.5 -16.492 -23.583 32.5 0

-0.16243 0.2936 0.3419 0.048818 22.845 58.5 16.5 -9.5427 -23.583 6 0

-0.10334 0.26443 0.39344 0.019838 6.9766 51.5 39.5 -9.5427 NaN NaN 0

-0.43722 0.25071 0.30092 0.12248 -1.2509 37.5 17.5 -9.5427 -8.965 32.5 0

-0.64473 0.16058 0.28473 0.32896 -1.2509 12 21 -16.492 16.562 50 0

-0.47479 0.32879 0.36768 0.29809 -1.2509 23.5 39.5 9.8324 16.562 25 0

-0.43228 0.38063 0.4039 0.25806 -1.2509 25.5 68 9.8324 16.562 19.5 0

-0.47773 0.36195 0.36284 0.17181 -9.9903 32 100.5 24.033 24.472 43.5 0

-0.5734 0.22027 0.27696 0.22512 -1.2509 39.5 77 9.8324 16.562 19 0

-0.59505 0.16465 0.25177 0.032086 -1.2509 20.5 49 -9.5427 16.562 56.5 0

-0.52266 0.26686 0.25899 0.23961 -1.2509 14.5 25.5 -9.5427 -23.583 14.5 0

-0.53232 0.36506 0.2831 0.276 -9.9903 66.5 39.5 -9.5427 -23.583 38.5 0

-0.51283 0.41519 0.26584 0.31275 -31.993 48.5 21.5 -9.5427 -23.583 39 0

-0.52028 0.22885 0.20798 0.24116 -1.2509 34 30 -9.5427 -23.583 9 0

-0.64073 -0.23303 0.3239 0.096257 -1.2509 30.5 14 -16.492 -23.583 15.5 0

-0.58313 -0.47048 0.21717 0.61101 6.9766 22 17.5 -16.492 16.562 42 0

-0.66566 -0.60853 0.32624 0.50543 6.9766 39 20 -16.492 16.562 52 0

-0.72746 -0.70762 0.42629 1 -1.2509 29.5 17 -9.5427 -23.583 49.5 0

-0.88344 -0.74499 -0.066155 0.90357 -1.2509 28.5 23.5 -9.5427 16.562 31.5 0

-0.89546 -0.67987 0.018943 0.94204 -1.2509 20.5 25.5 -16.492 -8.965 61.5 0

[Ive J]

I could not reproduce your error for both datasets you mentioned:

head(t1)
     ThetaX     ThetaY     ThetaZ       Td         Vd        Vw      Hw       Hd         Ld       Lw 
    ________    _______    _______    _______    _______    ____    ____    _______    ______    ____

      1.2333     1.2445    0.52498    0.53232     6.5127    34.5    71.5    -15.891       NaN     NaN
      1.6859     2.1313    0.93286    0.90519     6.5127    37.5     NaN        NaN       NaN     NaN
      1.3092     0.5145     3.7508    0.58472     29.102      41      55    -15.891       NaN     NaN
    -0.52485     1.4835     4.1045          1     29.102      37     NaN        NaN       NaN     NaN
     -11.536    -4.3153     13.498          0     36.438      38     NaN        NaN       NaN     NaN
     -4.6048    -3.5271     7.2064    0.55741     14.538    41.5     NaN        NaN       NaN     NaN
      13.477    -2.3554     2.5966          0     -1.105      39      35     15.911    15.182    14.5
      26.936    -4.7887     5.1617    0.84217    -18.806      40      39     15.911    15.182    37.5

mdl1 = fitglme(t1,'Vw ~ Td + (1|ThetaX) + (1|ThetaY) + (1|ThetaZ)','FitMethod','Laplace'); % no error
mdl1
Generalized linear mixed-effects model fit by ML

Model information:
    Number of observations              22
    Fixed effects coefficients           2
    Random effects coefficients         65
    Covariance parameters                4
    Distribution                    Normal
    Link                            Identity
    FitMethod                       Laplace

Formula:
    Vw ~ 1 + Td + (1 | ThetaX) + (1 | ThetaY) + (1 | ThetaZ)

Model fit statistics:
    AIC       BIC       LogLikelihood    Deviance
    133.39    139.94    -60.696          121.39  

Fixed effects coefficients (95% CIs):
    Name                   Estimate    SE        tStat       DF
    {'(Intercept)'}           38.1     2.2593      16.864    20
    {'Td'         }        -3.1335     3.2122    -0.97551    20


    pValue        Lower     Upper 
    2.7251e-13    33.388    42.813
       0.34096    -9.834     3.567

Random effects covariance parameters:
Group: ThetaX (22 Levels)
    Name1                  Name2                  Type       
    {'(Intercept)'}        {'(Intercept)'}        {'std'}    


    Estimate
    0.67125 

Group: ThetaY (22 Levels)
    Name1                  Name2                  Type       
    {'(Intercept)'}        {'(Intercept)'}        {'std'}    


    Estimate
    0.67125 

Group: ThetaZ (21 Levels)
    Name1                  Name2                  Type       
    {'(Intercept)'}        {'(Intercept)'}        {'std'}    


    Estimate
    3.7062  

Group: Error
    Name                        Estimate
    {'sqrt(Dispersion)'}        0.98473 

also for second dataset:

head(t2)
 ThetaX       ThetaY      ThetaZ        Td         Vd       Vw      Hw       Hd         Ld       Lw     Ph
    _________    ________    ________    ________    ______    ____    ____    _______    ______    ____    __

       1.1022     -1.0173    0.067847     0.60151    31.072      45    60.5    -16.492       NaN     NaN    0 
       1.2061     -1.0606     0.16396           1    15.428    44.5    62.5    -16.492       NaN     NaN    0 
      0.97804     -1.0005     0.19531     0.27491    6.9766    39.5    36.5    -16.492       NaN     NaN    0 
      0.76398    -0.97725      0.1889           0    6.9766    41.5    12.5    -16.492       NaN     NaN    0 
    0.0093436     0.27112     0.17762    0.099879    6.9766    29.5    34.5    -16.492    24.472    59.5    0 
    -0.059496      0.4879     0.39032     0.21028    22.845      34    85.5    -16.492       NaN     NaN    0 
      0.12878     0.50104     0.34031     0.32137    31.072      12      14    -16.492       NaN     NaN    0 
    -0.040724     0.63091     0.40665     0.20183    6.9766    23.5    35.5    -16.492       NaN     NaN    0 
 
mdl2 = fitglme(t2,'Vw ~ Td + (1|ThetaX) + (1|ThetaY) + (1|ThetaZ)','FitMethod','Laplace'); % no error
mdl2
Generalized linear mixed-effects model fit by ML

Model information:
    Number of observations              30
    Fixed effects coefficients           2
    Random effects coefficients         90
    Covariance parameters                4
    Distribution                    Normal
    Link                            Identity
    FitMethod                       Laplace

Formula:
    Vw ~ 1 + Td + (1 | ThetaX) + (1 | ThetaY) + (1 | ThetaZ)

Model fit statistics:
    AIC       BIC       LogLikelihood    Deviance
    253.16    261.56    -120.58          241.16  

Fixed effects coefficients (95% CIs):
    Name                   Estimate    SE        tStat       DF
    {'(Intercept)'}         32.695     3.6806       8.883    28
    {'Td'         }        -2.3505     8.4533    -0.27805    28


    pValue        Lower      Upper 
    1.2288e-09     25.155    40.234
       0.78301    -19.666    14.965

Random effects covariance parameters:
Group: ThetaX (30 Levels)
    Name1                  Name2                  Type       
    {'(Intercept)'}        {'(Intercept)'}        {'std'}    


    Estimate
    6.7341  

Group: ThetaY (30 Levels)
    Name1                  Name2                  Type       
    {'(Intercept)'}        {'(Intercept)'}        {'std'}    


    Estimate
    6.7341  

Group: ThetaZ (30 Levels)
    Name1                  Name2                  Type       
    {'(Intercept)'}        {'(Intercept)'}        {'std'}    


    Estimate
    6.7341  

Group: Error
    Name                        Estimate
    {'sqrt(Dispersion)'}        6.7341  

[Sam Litvin]

Thanks Ive, I only included part of the data sets. I found that when I exclude some of the data it works. It seems to be connected with some of the subsets of the data. When I excluded it, it worked. I noticed it was tied to number of NaN's in a data set.