Docsity
Docsity

Prepare for your exams
Prepare for your exams

Study with the several resources on Docsity


Earn points to download
Earn points to download

Earn points by helping other students or get them with a premium plan


Guidelines and tips
Guidelines and tips

Nonresponse - Survey Sampling Techniques - Lecture Slides, Slides of Survey Sampling Techniques

Survey Sampling Techniques course is one of important courses in Statisitics. Major poiuts of this course are: probability sampling, confidence intervals, Two-stage cluster sampling, Two-stage cluster sampling, estimation for mean, choosing strata, allocation across strata, ratio estimation, domain estimation, Two-stage cluster sampling. Keywords in these slides are: Nonresponse, Design Phase, Weighting Adjustments, Imputation of Missing Values, Eligibility Known, Unit Nonresponse, Item Nonresp

Typology: Slides

2012/2013

Uploaded on 08/30/2013

faroq
faroq 🇮🇳

4.1

(14)

104 documents

1 / 36

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Nonresponse
Whatisnonresponse(NR)?
ImpactofNRpotentialbiasandlossofprecision
StrategiestoreduceNR
Designphase
Afterdatacollection
Secondsurveytocollectdataonnonrespondents(double
sampling)
Weightingadjustments
Imputationofmissingvalues
docsity.com
pf3
pf4
pf5
pf8
pf9
pfa
pfd
pfe
pff
pf12
pf13
pf14
pf15
pf16
pf17
pf18
pf19
pf1a
pf1b
pf1c
pf1d
pf1e
pf1f
pf20
pf21
pf22
pf23
pf24

Partial preview of the text

Download Nonresponse - Survey Sampling Techniques - Lecture Slides and more Slides Survey Sampling Techniques in PDF only on Docsity!

Nonresponse

-^

What

is

nonresponse

(NR)?

-^

Impact

of

NR

potential

bias

and

loss

of

precision

-^

Strategies

to

reduce

NR

–^

Design

phase

–^

After

data

collection

-^ Second

survey

to

collect

data

on

nonrespondents

(double

sampling) • Weighting

adjustments

-^ Imputation

of

missing

values

How

nonresponse

occurs

•^

Can

not

locate/contact

SU

-^

May

not

know

if

SU

is

eligible

•^

Contact

SU,

but

SU

refuses

to

participate

-^

May

or

may

not

know

eligibility

of

SU

•^

SU

agrees

to

participate,

but

doesn’t

answer

all

questions

-^

Eligibility

known

Unit

nonresponse

-^

SRSWOR

of

n

from

N

Sus

-^

Of

the

n

SUs,

only

n

respond R

–^

n^ R

n

-^

Simple

response

rate

–^

Proportion

-^

Percentage

-^

Simple

nonresponse

rate

R M^

nn n^



nR^ n nR^ n^100

nM n

Nonresponse

population

framework

Whole Population

N^

elements with pop mean

1

 ^

Ui

i

U^

y N y

Nonresponse

framework

and

population

parameters

Nonresponse

population

framework

-^

Relationship

between

population

mean

and

means

for

responding

and

nonresponding

subpopulations

RU

RU

MU M

RU M

MU M

RU R

MU M

U

y

y

y N N

y N N

y N N

y N N

y N N y

 

)

(

)

(^1) (

RU R MU M

Ui R

i R

Ui M

i M

Ui

Ui

i

i

Ui

i

U^

y N N y N N

N

y N

N

y N N y y N y N

y^

R

M

M^

R

      

^

^

^

)

1 ( )

1 (

1

Derivation

Nonresponse

sample

framework

Nonrespondents

(M)

Respondents

(R)

Sample

n =

nM

nR

elements

mean ent

nonrespond

onn

informatio no

???

respond not will

units

sample^  M nM y

s

respondent

of

mean

(^1) sample

respond will SUs^ 

^ AiR

i

R R R

y

n n y

Nonresponse

bias

-^

Nonresponse

bias

occurs

when

differences

exist

between^ –

the

population

mean

of

y^

for

the

nonresponding

subpopulation

and

–^

the

population

mean

of

y^

for

the

responding

subpopulation

-^

The

magnitude

of

nonresponse

bias

depends

on

–^

Difference

between

population

means,

–^

Nonesponse

rate,

n^ M

/^

n^

=^

n^ R

/^

n

y^ MU yRU

RU

MU

y

y^

Strategy

Design

to

prevent

-^

Consider

likely

mechanisms

for

NR

when

designing

survey

-^

Reduce

respondent

burden

to

extent

possible

-^

Main

areas

-^

Survey

setting

and

timing

-^

Data

collection

methodology

-^ Burden

for

individual

-^

Sample

design

-^ Burden

for

population

-^

Remedies

for

avoiding

NR

also

tend

to

improve

data

quality

-^

Read

more

in

Chapter

8

of^

Lohr.

Strategy

2:

Use

data

from

call

‐backs

of

NR

cases

to

adjust

for

bias

-^

Basic

idea

–^

Select

a^

sample

from

the

nonrepsondents

to

the

survey

–^

Collect

data

from

contacted

nonrespondents

–^

Use

these

data

to

estimate

population

mean

for

nonrespondents

-^

Estimate

population

mean

for

whole

pop

with

a^

weighted

combination

of

respondent

sample

mean

and

nonrespondent

sample

mean

•^

This

sample

design

is

an

example

of

“double”

or

‐phase”

sampling

(we

won’t

cover

this

in

y^ MU

y^ U

Estimation

approach

•^

Sample

mean

from

responding

population

-^

Sample

mean

from

“call

‐back”

subset

of

nonresponding

population

(assume

response)

-^

Unbiased

estimator

for

population

mean

–^

weighted

mean

of

two

sample

means

n^ R i

i R R

y n y^

1 1

n^ MCB i

i

MCB M

y

n y^

1 1

M M R R CB^

y n n y n n y^

 ˆ

Estimation

approach

•^

Estimator

for

variance

•^

Assumptions^ –

SRSWOR

at

both

phases

of

sampling

-^

All

call

‐back

respondents

provide

data

response

on

subsample)

 

^ 

 

  

  ^

2

2

2

2

) ˆ ( ) ˆ (

1 1

(^11)

(^11)

) ˆ( ˆ^ V

CB M M

CB R R M M R R

CB^

y y n n

y y n n

n s n

n n s n

n n

y

docsity.com

Basic

problem

for

unit

nonresponse

-^

In

general,

sample

may

not

reflect

composition

of

population

-^

Especially

with

SRS

-^

With

nonresponse,

even

if

we

start

with

a

“representative”

sample,

we

can

get

a

set

of

respondents

sample

that

do

not

reflect

the

composition

of

the

population

Basic

problem

for

unit

nonresponse

•^

We

will

get

nonresponse

bias

from

an

unadjusted

sample

mean

of

respondents

if^

our

analysis

variable

is

related

to

the

nonresponse

mechanism

-^

Suppose

we

are

studying

a^ disease

that

is^

more

likely

to

occur

in

males

-^

y = j^

1 if

person

j^ has

the

disease,

0 otherwise

-^

Suppose

females

tend

to

respond

at

a^ higher

rate,

but

they

get

the

disease

less

often

-^

Using

a^ simple

mean

estimator

is^

likely

to

underestimate

the

true

disease

prevalence

in

the

population

-^

Remedy

(gender

as

groups,

or

post

‐strata):

make

an

estimate

for

males

and

an

estimate

for

females

and

pool

using

the

known

proportion

of

males

and

females

in^

the

target

population