Proportions
Proportions
Strong Interest in fish/wildlife
Examples?
Proportions
\[ y \in \{0,1\} \] \[ 1 = \text{occur} \\0 = \text{not occur} \]
. . .
\[ \hat{p} = \frac{1}{n}\sum_{i=1}^{n}y_i \]
. . .
\[ \hat{\sigma}^2_{p} = \left(\frac{N-n}{N}\right)\frac{\hat{p}(1-\hat{p})}{n-1} \]
Proportion area occupied
How does sample grid effect \(\hat{p}\) and \(\hat{\sigma}^2_{p}\)?
Proportion area occupied
How does sample grid effect \(\hat{p}\) and \(\hat{\sigma}^2_{p}\)?
Proportion area occupied
How does sample grid effect \(\hat{p}\) and \(\hat{\sigma}^2_{p}\)?
Var(p)
. . .
SD(p)
Soooooo useful!
\[ \hat{\sigma}_{p} = \sqrt\frac{\hat{p}(1-\hat{p})}{n-1} \]
. . .
- how many fish should I tag?
- how many elk should I track?
- how many point intercepts (plant survey) should I take?
- how many plots (camera traps / plant survey) should I sample?
SD(p)
Lets visualize a few standard deviations of p
n, p, sd(p)
My Wolverine ‘Problem’
A Sampling ‘Problem’
- Sites surveyed by the States were selected via SRS
- a handful of additional sites were surveyed by other entities (e.g., NGO) using the same field protocol
. . .
Should I include the additional/ancillary surveys as data?
Should I report the cells with occurrences at all?
. . .
In small groups: Discuss and then tell me what to do and explain.
Simulation
What if?
- N = 600
- p = 0.5
- SRS of n = 200
sites occu
1 1 1
2 2 1
3 3 0
4 4 0
5 5 0
6 6 1
SRS
Sampling Distribution of \(\hat{p}\)
Estimator Bias = -0.00005
Distribution Var = 0.00084
Ancillary Occurences
SRS of n = 200 with 5% ancillary and occupied sites
Estimator Bias = 0.02382
Distribution Var = 0.00073
Ancillary Occurences
SRS of n = 200 with 10% ancillary and occupied sites
Estimator Bias = 0.04536
Distribution Var = 0.00069
My Wolverine ‘Problem’
Should I include the ancillary surveys as data?
Should I report the cells with occurrences at all?
Unequal Probability of Sampling
Instead of adding sites w/o clear sampling probabilities, we make these explicit.
Identify sites that are important and increase their probability of being selected for sampling
Unequal Probability of Sampling
1/2 sites with 3x the probability of being sampled versus other half
sample
function has an argumentprob
. . .
\[ \hat{p} = \frac{1}{n}\sum_{i=1}^{n}y_i \]
. . .
Estimator Bias = 0.13753
Unequal Probability of Sampling
Generalized Unequal Probability Estimator
Thompson Ch. 6.3
\[ \hat{\theta} = \frac{\sum_{i} y_i / \pi_i}{\sum_{i}1 / \pi_i} \]
- \(\pi_{i}\) = probability of selecting sample \(i\).
. . .
Estimator Bias = -0.03147
“not precisely unbiased” - Thompson