1

I have some data with predictors that work out the value of changes applied to an ID.

Each TYPE is a predictor for the CHANGE applied to an ID.

Below is an example of the data:

ID  CHANGE   TYPE    VALUE   STATE
A   RA       1       1.14    POS
A   RA       2       0.96    POS
A   RA       3       0.78    POS
A   RA       4       0.84    POS
A   RA       5       0.92    POS
B   SR       1       1       POS
B   SR       2       0.82    POS
B   SR       3       -0.012  NEG
B   SR       4       -0.036  NEG
B   SR       5       -0.02   NEG
C   GW       1       -0.86   NEG
C   GW       2       -0.77   NEG
C   GW       3       -0.82   NEG
C   GW       4       -0.62   NEG
C   GW       5       -0.74   NEG

Is there a measure of a distance/spread I could apply to each ID? For example, C has all predictions very close to each other so we can be confident in the prediction but B's predictions are spread out so we are less confident.

data example

DataD
  • 11
  • 1
  • 4

1 Answers1

1

Perhaps too obvious, but you could use the standard deviation of the predictions, or a confidence interval if you have enough predictions.

Firebug
  • 19,076
  • 6
  • 77
  • 139
  • Apologies for the delay, thank you for answering. I was thinking about standard deviation or the mean absolute deviation :)

    I wasn't sure about confidence intervals as the points aren't a sample from a population?

     – DataD Aug 18 '16 at 14:57