Attempt to resolve warnings in effect_N_discrete #97
Conversation
| warning('Extrapolating the rarefaction curve because the number of rescaled individuals is smaller than the inds argument') | ||
| interp_S_samp = pchip(c(1, rescaled_effort), c(1, S_samp), inds) | ||
| # No extrapolation of the rescaled rarefaction curve, only interpolation | ||
| interp_S_samp = pchip(c(1, rescaled_effort), c(1, S_samp), inds[inds <= max(rescaled_effort)])[1:length(inds)] |
There was a problem hiding this comment.
Nice job fixing this! I can see why this code works but it will difficult to understand for future us and other users. If we don't know that R adds NA's to cells that are longer than the vector is defined across we would be scratching our heads right now. I would prefer one step where you define interp_S_samp and then a second step where you add NA's to that vector so that it has the same length as inds. What do you think?
| ind_sample_size) | ||
| effect_N_perm[, j] = group_effect_N_perm$deltaS | ||
| ind_sample_size[ind_sample_size <= sum(comm_level_perm)]) | ||
| effect_N_perm[, j] = group_effect_N_perm$deltaS[1:nrow(effect_N_perm)] |
There was a problem hiding this comment.
I have the same basic feeling here as in the previous comment.
| ind_sample_size)$deltaS) | ||
| null_N_deltaS_mat[i, ] = N_eff_perm[ , 2] - N_eff_perm[ , 1] | ||
| ind_sample_size[ind_sample_size <= min_N])$deltaS) | ||
| null_N_deltaS_mat[i, ] = (N_eff_perm[ , 2] - N_eff_perm[ , 1])[1:ncol(null_N_deltaS_mat)] | ||
| setTxtProgressBar(pb, k) |
There was a problem hiding this comment.
and here, but great job diving in and tracking these problems down. I'll test your code out on the case-studies to see if anything changes drastically on the results.
|
Sure & done! (Except for the second case, which is already on its own line. But I did add a note to our future selves.) |
|
Ok yea this isn't quite was I was thinking something more like |
This issue is described in #21 .
There were two warnings:
rarefactionoccurred when the summed N of two treatments were close. By swapping plot-level n among plots, the treatment with smaller N to start with would sometimes end up with even lower N, and as a result the individual-based curve couldn't be built all the way to the previous ending point.deltaS_Noccurred when the treatments differed considerably in average density. The rarefaction curve of the higher-density-treatment could be shrunk below the ending point by the rescaling indeltaS_N.In both cases I implemented the fix to fill the dataframe with
NA'sat the bottom, instead of rarefying beyond totalNor extrapolating the rarefaction curve beyond its ending point.