Prev: last column of *.prn file missing
Next: proc report and generate frequencies based on a number of
From: Jeff Nielsen on 24 Feb 2010 15:14
I am looking for the best proc to analyize non-uniform ordinal data.

The trial is a randomized block design with 4 replications.

The treatments are 4 different varieties of sugarbeet each having
differing resistance to the disease.

At the time of harvest we sub-sample 20 roots and give them a rating based
on the following scale.

0= clean no visiable lesions
1= Superficial scattered non active lesions
2= active lesions affecting < 5% of the root
3= extensive lesions affecting 6-25% of the root
4= extensive rot 26-50% of the upper half of the root with cankers
5= > 50% of the root blackend with rot to the interior of root
6= entire root blackened except the tip
7= 100% rotted and the plant is dead

As you can see by the scale there is a great difference between the
ratings that does not carry over to a simple percentage conversion.

historically we take an average for the 20 roots and come up with a mean
for the plot. Should we be analizing all 20 roots, I dont know the answer.

All help and suggestions appreciated

thanks
Jeff
 | 
Pages: 1
Prev: last column of *.prn file missing
Next: proc report and generate frequencies based on a number of