Robertson Flashcards

1
Q

“Serious injuries”, 1993 NCCI definition

A

Fatal

Permanent Total

Major Permanent Partial

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Non-serious injuries, 1993 NCCI definition

A

Minor permanent partial

Temporary total

Medical only

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

1993 NCCI hazard grouping

A

4 proposed hazard groups, linear combination of a few variables

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

L1 distance

A

Advantage: minimizes relative error in estimating excess premium

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

L2 distance

A

Euclidean distance; selected as analysis not sensitive to distance measure

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Why NCCI did not use standardization

A
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

NCCI standardizing considerations

A
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Non-hierarchical clustering

A

New hazard groups did not have to be subsets of existing hazard groups

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

k-means algorithm

A
  1. Decide on number of clusters k to target
  2. Start with initial arbitrary assignment into k clusters
  3. Compute centroid (mean) of cluster
  4. For each class, find closest centroid (L2 distance) and assign to that cluster

If there has been any movement, repeat step 3

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Tests to determine number of hazard groups

A

Calinski and Harabasz

Cubic Clustering Criterion (CCC)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Calinski and Harabasz statistic

A

“Pseudo-F test”

Higher values, better # of clusters

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Cubic Clustering Criterion

A

Compares amount of variance explained by a given set of clusters to that expected when clusters are formed at random; higher values better

Less reliable when data is highly correlated

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Why NCCI strayed from CCC

A
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

“Crossover”

A

Counterintuitive results (i.e. exceptions to rules)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Why NCCI went from 17 limits to 5 limits

A

ELFs at any pair of excess limits are highly correlated across classes

Limits below $100K were heavily represented

Wanted to cover range of limits commonly used for retro rating

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

5 limits in NCCI study (2007)

A

$100K

$250K

$500K

$1M

$5M

17
Q

Credibility formula for NCCI Hazard Mapping

A

Z = min (1.5n / (n + k) , 1)

n = # of claims in the class

k = average # of claims per class

18
Q

Other credibility options in NCCI Hazard Group Mapping

A
  1. Using median as k
  2. Excluding Med-Only claims
  3. Including only Serious claims
  4. Requiring minimum # of claims for classes used to calculate k
  5. Various square root rules
19
Q

UW feedback on proposed hazard grouping

A

Similarity between class codes in different hazard groups

Degree of exposure to auto accidents in a given class

Extent heavy machinery is used in a given class

20
Q

3 key ideas of remapping hazard groups

A
  1. Computing excess ratios by class
  2. Sorting classes based on excess ratios
  3. Cluster analysis