I am running a split-plot trial on SPSS. Everytime I run the Univariate analysis, one variable, "irrigation" (string, nominal variable) appears as a dot in the Tests Between Subjects table. I have tried redoing it many times and even checked my data but I always have the same problem.
I have included a screenshot of the problem as well
How many unique values do you have for irrigation? What is your sample size? And are there multiple values in each cell of the crosstabulation of the three between subject variables?
Daniel Wright I don't have any unique values, the sample size is 36. I'm not sure about the multiple values, I haven't considered that.
You should share a representative sample or whole of your data to check its structure.
Sorry, could you share also your code? And a frequency table of irrigation*zizanioktono*subplot ?
Mehmet Sinan Iyisoy You mean this code?
Also I'm not sure how to make a frequency table for all three variables, I can only make 3 separate ones, did you mean that?
Androniki Mina yes that's what I meant. I am not a regular user of SPSS, so I can't give the relevant frequency table code. I should find some SPSS installation.
I am sure @Bruce will respond in a short time.
I think that Mehmet Sinan Iyisoy is asking for a cross-tabulation of those three variables. This ought to give it:
CROSSTABS zizanioktono BY subplot BY irrigation.
Yes, that's it, although it would be easier to see without the row and column percentages included. Even so, one can see that when you cross-tabulate all of those variables, you have only one observation per cell.
Is that what's been causing the problem? If so should I change something or use a different code?
Try removing the interaction terms ... but in most contexts the answer is that you need more data. You have no variability/variance within cells, so Analysis of VARIANCE has difficulties.
I removed the interaction terms and tried it with "Main Effects" and "All 3-way" build terms, same result.
Androniki , Mina, Daniel Wright was correct there was a problem.
The sum of squares for the variables "
"irrigation" is 0 (zero)
For "zizanioktono" it is 78.810.
There is also the comment that the appropriate error term cannot be computed using Satterwaite's method.
Look at your output again.
Regards.
The 3-way is an interaction. Remove this too. I need to stress though, that depending on the research context (which I don't know), it is likely the sample is too small. If you get standard error estimates they will likely be large. You have too many cells. You can also only include one or two of the predictor/independent variables.
Androniki Mina,
1. Computer output comments are usually indicative of the possible source of problems. This is my own experience. If it happens that a zero is made to divide any value, the result is infinity. Of course to avoid an unending run time, the SPSS algorithm provides for replacing the expected result with stars. I know about this because I have used of the Fortran software for a long time.
2. Are the 3 variables being analysed in the same nominal scale units as "irrigation " variable is ?
Androniki Mina unfortunately you should explain more about your design factors. As far as I understand you have 3 different irrigation schemes and 4 different herbicides(zizanioktono). Is that correct? And if so, does subplots mean the block factor?
It seems that, as suggested above, you don't have enough variation or you don't have the correct code.
Harold Chike 3 of the variables are nominal (irrigation, zizanioktono and subplot), the other is scale
Mehmet Sinan Iyisoy yes the subplot is the block factor, as I mentioned I run a split-plot trial, and my main plots are the irrigation variations and subplos the zizanioktono.
I am aware that the sample is quite small, I am running this analysis as part of a simulation trial, it is not to say an actual field trial, but an assigniment for Biometry Class where we have to come up with a split-plot design trial and perform the analysis on it. So it might be in fact that there is a problem with my data and variables.
Hello, have you checked the variable irrigation, if declared as numeric, does not have strings in some hidden entry?
As I had indicated earlier. error messages from computer outputs are generally indicative of the sources of error. Out of your 3 variables, only 1,
"irrigation " has a problem with sprinting out. The other two variables have no problems. So go back and recheck your input data for irrigation.
Recheck whether the data on irrigation are all in the correct column space. The problem you are having may not be a statistical problem.
If it is , the other two variables will have the same problem. But that is not so. I happen to be an experienced computer programmer also.
- Badges
- Science topic