Search code examples
rstatisticssumconditional-statementspoisson

How can I conditionally sum given a latent indicator?

I have the following summation that I want to compute in R:

$\sum^K_{k=1} [(1-z_i) * log[y_i|\lambda_0] + z_i * log[y_i|\lambda_1]]$, where [y_i|\lambda_{1/2}] follows a Poisson(\lambda_{1/2}) distribution.

The latent indicator variable z_i is equal to either 0 or 1, which is provided in the vector z.

I currently have this code in my R script:

sum(dpois(y,lambda0,log=TRUE)[z==0]+dpois(y,lambda1,log=TRUE)[z==1])

I am receiving the following error,

longer object length is not a multiple of shorter object length

which I do not believe is really my issue in this case. I just do not know how to condition my summation. I am relatively new to R and would really appreciate your help.

Solution

  • If your y and z are of equal length, then you can try the code below, and I believe it is mathematically equivalent to your objective

    sum(dpois(y,lambda0,log=TRUE)*(1-z)+dpois(y,lambda1,log=TRUE)*z)