I was carrying out a study on perceptions of drought effects by farmers as influenced by their production environments. Production environments were my predictors and then I ranked drought impacts as scores. That is, respondents would score from 1 to 5 according to severity of impact: EFFECT Rank (1 to 5)
Water shortage
Feed shortage
Heat stress
Cattle mortality
My model was:
The effect of production environment on number of different classes of cattle lost to drought and rank scores for breed preferences, challenges to cattle production, reasons for loss of cattle, severity of drought components, severity of the effects of drought components on cattle production, grazing management problems during drought, types of parasite prevalent during droughts and breed susceptibility to drought were determined using PROC GLM SAS (2008). The model used was:
Yij=µ +EiI+ ɛij
Where:
Yij= response variable (number of cattle lost to drought; rank scores for: severity of the effects of drought attributes on cattle production, severity of the effects of drought on cattle, grazing management problems during drought, types of parasite prevalent during droughts and breed susceptibility to drought)
µ = mean common to all observations;
Ei= effect of production environment (sub-humid areas ; semi-arid areas);
ɛij= residual error ~ N (0; Iσ2).
[MC1]Sounds wrong…some variables are discontinuous.
Hi Takunda Dzavo, Thanks a lot for conducting research on drought stress, I think you use GLM to analyze ranked data.
You are trying to model an ordinal variable, so you should use ordinal regression: https://en.wikipedia.org/wiki/Ordinal_regression#:~:text=In%20statistics%2C%20ordinal%20regression%20(also,between%20different%20values%20is%20significant.
Hi Takunda, you can but you should not. The general linear model is for continuous data, you should analyze it was multivariate or you can set the response of yes (1) or not (0) to each rank and then present the % in each rank and you can use a GLIMIX with a binomial distribution.
in sas:
proc glimmix;
class block trt ;
model rank1 = trt block/ dist=binomial ddfm=satterth;
lsmeans trt / pdiff ilink;
run;
good luck
ale
- Badges
- Science topic
- Similar topics
- Philosophy
- Philosophy Of Science
- Reasoning