Results Interpretation

#1
Hi,

I am new here.. I mean new to statistics. In my study, I had to use some statistical analysis (logistic regression and discriminant analysis). My question is if I have a working model with good accuracy to predict an action (pass/fail) how do I say given a set of variable values, this student will fail or pass? (I dont know the final interpretation of the results). I just want to know which part of the SPSS output should I be looking on to be able to say that?

Am I making sense? Please help. If possible please dont use too much statistical terms as I am a noobie here.
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
How are you judging accuracy? How many variables do you have in the model and how are they formatted? What is you sample size?

You will likely need to convert your log odds results into probabilities or present as odds ratios. However, if you have multiple covariates, you will have multiple probabilities and ORs.
 
#3
How are you judging accuracy? How many variables do you have in the model and how are they formatted? What is you sample size?

You will likely need to convert your log odds results into probabilities or present as odds ratios. However, if you have multiple covariates, you will have multiple probabilities and ORs.
Hi hlsmith,
Thanks for replying. Actually what I said was done by a researcher and she hasn't mentioned how she arrived at the conclusion. I am trying to making sense of it. She has 7 variables and around 120 sample size. They are all quantitative if thats what you mean. She says the model has 88% accuracy and I am guessing that is from the SPSS output. Is there a way to do so?

The hypothesis was to identify the files created on a particular folder on a computer is the result of an intentional or unintentional action. So the need is to suggest that given the values of the variables, we should be able to tell if it is intentional or unintentional.

Any suggestions on how to arrive at that result from the model?

Thanks
Stephin
 

hlsmith

Less is more. Stay pure. Stay poor.
#4
Accuracy in logistic regression is typically measured by the concordance statistic (c-statistic), known as the area under the curve in some settings. The continuous variable can be used to find a cutoff value or will be interpreted as increased odds given a unit increase in the variable.
 
#5
Accuracy in logistic regression is typically measured by the concordance statistic (c-statistic), known as the area under the curve in some settings. The continuous variable can be used to find a cutoff value or will be interpreted as increased odds given a unit increase in the variable.
How about measuring accuracy in discriminant analysis? Is there an easy way?

Thanks
Stephin