由买买提看人间百态

boards

本页内容为未名空间相应帖子的节选和存档,一周内的贴子最多显示50字,超过一周显示500字 访问原贴
Statistics版 - approval rate report
相关主题
[合集] 怎么算AUC[合集] 请教PROC MIXED 高手:find difference in mean change f
How to select baseline(s) for acquisition score performance report?问一个underpower和predictive probability的问题
急问一个analysis of covariance的问题请教一个longitudinal data的分析问题
question on time-dependent baseline in SAS phreg!菜鸟请问treatment在control/mutant的作用用什么test
A question on linear regression发面经,攒RP期望background check能过
如何对这个数据做统计模型?Help: mixed model
问两个个KNN的问题SAS小白问个问题好么?
logistic regression: baseline vs. (baseline + additional variable)菜鸟请问ANOVA问题
相关话题的讨论汇总
话题: rate话题: baseline话题: approval话题: accept话题: psi
进入Statistics版参与讨论
1 (共1页)
y*****n
发帖数: 5016
1
anyone working on acquisition model performance monitoring reports?
i am looking for a template of approval rate report. however, all that i
have found is a table with the below columns:
Score Applications# Accepts# Declines# Accept Rate (Bin) Accept
Rate (Cum)
obvious, this is not sufficent, i expect at least some other columns for
baseline to compare with and at least a column that measure the different
between current month/quarter and the baseline.
can anyone provide me a complete template? especially i would like to know
how you measure the different between current month/quarter and the baseline
-- the same way as you calculate PSI for all applications or a different
way?
thank you!
c*********y
发帖数: 122
2
Not sure your purpose.
Firstly, there is seasonality and credit policy fine-tune/change. Not sure
you can keep a fixed baseline for a long period time.
Secondly, your marketing department might fool you by targeting more peoples
below your cut-off which might cause higher applications than usual but low
approval rate. So it is hard to set a baseline.
Just my 2 cents.
j**********3
发帖数: 305
3
baseline will be developmental sample. Decile the dev sample by score. Then
apply the same scoreband on your recent thru the door sample. Base on cum
volume for dev sample and recent sample. You can calculate KS and PSI. Then
you apply the same method on each individual characteristic in ur Acq model
.

Accept
baseline

【在 y*****n 的大作中提到】
: anyone working on acquisition model performance monitoring reports?
: i am looking for a template of approval rate report. however, all that i
: have found is a table with the below columns:
: Score Applications# Accepts# Declines# Accept Rate (Bin) Accept
: Rate (Cum)
: obvious, this is not sufficent, i expect at least some other columns for
: baseline to compare with and at least a column that measure the different
: between current month/quarter and the baseline.
: can anyone provide me a complete template? especially i would like to know
: how you measure the different between current month/quarter and the baseline

y*****n
发帖数: 5016
4
of course the approval rate report will be reviewed along with override rate
report otherwise it will be meaningless.
baselines are not difficult either. i think i can choose the first scoring
as well as prior month scoring as baselines.
the most difficult part is how to measure the change. i don't think the PSI
formula can apply to approval rate. PSI involves the application # column
only, while approval rate involves application # AND appr #.

peoples
low

【在 c*********y 的大作中提到】
: Not sure your purpose.
: Firstly, there is seasonality and credit policy fine-tune/change. Not sure
: you can keep a fixed baseline for a long period time.
: Secondly, your marketing department might fool you by targeting more peoples
: below your cut-off which might cause higher applications than usual but low
: approval rate. So it is hard to set a baseline.
: Just my 2 cents.

c*********y
发帖数: 122
5
Again, what's your purpose?
The purpose of PSI is to identify whether your underlying scored population
shift over time and which bucket of the shift.
Then you replace it with your approval rates by score buckets. What do you
want to tell your boss? opportunities loss?
y*****n
发帖数: 5016
6
you may want to do some homework by finding OCC's requirements on model
performance monitoring reports. and i fact, i have somehow answer your
question by mentioning that the approval report will be reviewed along with
the override rate report.

population

【在 c*********y 的大作中提到】
: Again, what's your purpose?
: The purpose of PSI is to identify whether your underlying scored population
: shift over time and which bucket of the shift.
: Then you replace it with your approval rates by score buckets. What do you
: want to tell your boss? opportunities loss?

C******t
发帖数: 72
7
You may use AUC or Sommer'd to track the model performance change.
s*********e
发帖数: 1051
8
this could be tricky, since some declines are by policy instead of score

Accept
baseline

【在 y*****n 的大作中提到】
: anyone working on acquisition model performance monitoring reports?
: i am looking for a template of approval rate report. however, all that i
: have found is a table with the below columns:
: Score Applications# Accepts# Declines# Accept Rate (Bin) Accept
: Rate (Cum)
: obvious, this is not sufficent, i expect at least some other columns for
: baseline to compare with and at least a column that measure the different
: between current month/quarter and the baseline.
: can anyone provide me a complete template? especially i would like to know
: how you measure the different between current month/quarter and the baseline

y*****n
发帖数: 5016
9
In my vision, models and strategies are always inseparatable. therefore,
definately yes i will cross-check with override rate report as have
mentioned in my previous threads.

【在 s*********e 的大作中提到】
: this could be tricky, since some declines are by policy instead of score
:
: Accept
: baseline

1 (共1页)
进入Statistics版参与讨论
相关主题
菜鸟请问ANOVA问题A question on linear regression
Intent to treat population如何对这个数据做统计模型?
求助,如何解决这个题目,5个包子答谢解决者!!!问两个个KNN的问题
"Missing data" "intent-to-treat" "repeated measure"logistic regression: baseline vs. (baseline + additional variable)
[合集] 怎么算AUC[合集] 请教PROC MIXED 高手:find difference in mean change f
How to select baseline(s) for acquisition score performance report?问一个underpower和predictive probability的问题
急问一个analysis of covariance的问题请教一个longitudinal data的分析问题
question on time-dependent baseline in SAS phreg!菜鸟请问treatment在control/mutant的作用用什么test
相关话题的讨论汇总
话题: rate话题: baseline话题: approval话题: accept话题: psi