ERROR: Invalid or missing data in Proc Glimmix

[s-k]

I am using the SAS OnDemand for Academics (free online version), I have attached a portion of my code. The only error code I am getting is that there is missing or invalid data but I'm not sure where that would be coming from. It will also give a NOTE: Invalid data for DAP in line 144 1-18. which repeats for  YEAR in line 145 1-18, PLOT in line 146 1-17, REP in line 147 1-17, TRT in line 148 1-17, NDVI in line 149 1-17  and this pattern repeats through the code. Any advice would be welcome!

data alldata;
input DAP YEAR PLOT REP TRT NDVI;
datalines;
114 1 402 4 3 0
114 1 403 4 2 0
114 1 404 4 1 0
125 2 401 4 2 0.855
125 2 402 4 3 0.825
125 2 403 4 4 0.815
119 3 303 3 3 0.69
119 3 304 3 2 0.775
119 3 403 4 4 0.705
;
run;
proc sort data=alldata; BY Year DAP;
proc glimmix data = alldata; BY Year DAP;
class REP TRT;
model NDVI = TRT;
random REP;
lsmeans TRT / lines;
run;

[tom_grant]

I had no trouble running your code in SAS Ondemand - I did get the following warning messages in the log for proc glimmix:

WARNING: Mixed model has saturated mean and profiled variance. Fit does not proceed.

NOTE: The above message was for the following BY group:

YEAR=1 DAP=114

WARNING: Mixed model has saturated mean and profiled variance. Fit does not proceed.

NOTE: The above message was for the following BY group:

YEAR=2 DAP=125

WARNING: Mixed model has saturated mean and profiled variance. Fit does not proceed.

NOTE: The above message was for the following BY group:

YEAR=3 DAP=119

[s-k]

Thank you for running it, I had to make sure it wasn't a code issue before going further! It turns out that in the free SAS OnDemand, the tabs/cells when copied over from any sort of sheet is not recognized as spaces and the program reads it as one long line of code. 

[JackieJ_SAS]

My first suggestion is to remove this statement from within the PROC GLIMMIX code:

BY Year DAP;

This fits a separate model for every unique combination of Year and DAP. The data you provided has three unique combinations of Year and DAP with three observations within each combination. So- you are fitting three separate models, each estimated with only three observations. Three observations isn't enough to fit a model well.

[Ksharp]

Your data is too small that would lead to a sparse problem for your model .

BY Year DAP;
would split your dataset into many sub-dataset.


random REP;
that would take REP as a SUBJECT= variable(cluster variable).

All of these is to say you don't have enough data to buid a Mixed Model.