NR449 Mod4 Nov24

hey everybody so let's talk this week

about sampling strategy to begin with

beginning with chapter seven in your

book it's also very well outlined in

your edep I hope you're paying attention

to that

okay so we're going to be talking about

sampling strategies sampling error and a

lot you know we talked we got into the

conversation last week about qualitative

versus quantitative data and so we're

going to continue on with qualitative

and quantitative research looking at

differences and comparing and whatnot

and so let's get

started one of the things that you

always need to be concerned about with

with research is bias bias of your own

bias the way the study is laid out

there's a lot of different ways to to

look for to have

bias one of the ways is we pick our

subjects based on how we think they're

going to respond I want to know that my

research that my way of doing things is

better than the old way so I may pick my

sample based on how I think they're

going to respond and not randomly or

impartially so I've biased my results

and they may not be

right which is going to threaten the

validity of the study certainly going to

uh affect the usability of the study we

control that like I said through random

sampling random assignment to groups so

if I have a group of a

thousand people that have

cancer and I decide that I want a sample

of 500 so I will throw a thousand names

in the head and I'll pull out until I

have my 500 randomly decided randomly

chosen I pulled out 500 I could use a

table of random numbers and do the same

thing where I could put everybody's

number give everybody a number and then

pull numbers randomly until I have 500

so once I've done that then I take my

500 and I'll randomly assign them to a

group as well so maybe I'm looking at

two different things and I will randomly

decide who goes to which group maybe

this is the control group and this is

the group that actually has is going to

get the treatment we did that with study

of single cell patients when we started

penicillin prophylaxis we wanted to know

in our children if penicillin

prophylaxis would prevent death would

prevent hospitalization and serious

complications of infection and so we

started a study and we put everybody

when they came to Clinic they were asked

to participate and then those that

participated were randomly assigned to

either the treatment group they got

penicillin or to the control group and

they got vitamin C neither the patient

nor the physician nor the nurses knew

what that patient was getting so we

didn't know whether they had penicillin

or not and that's called Blind we did

blindly assign them to groups

randomly sampling size

it is really what you're looking at to

determine that you have enough subjects

from that

population to get valid results to get

quality results two

people all the subjects agreed right but

you only had

two I have a thousand and 980 agreed now

you've got much more data to go with

right so you want it to represent the

population too you don't want to pick

and choose that bias

but you want to have a group when you're

done of participants that really reflect

what's out

there so how do you choose them well

first of all you define that population

Define exactly what you want maybe

patients between the age of 20 and 40

who have type 2

diabetes we could say they're overweight

or not you could give you could take a a

gender or not but Define that population

that you want to do and then look at

those members of the population who are

attainable your sampling frame who do

you have access

to finally look at inclusion and

exclusion criteria we're going to talk a

little more about that but who are you

going to include and who are you going

to exclude it's really important and why

do you exclude some is that going to buy

us your study

then so inclusion and exclusion criteria

really are the demographics maybe what

are we going to

include children between the age of five

and 10 who live in a certain area who

are able to read and write and complete

the questionnaire who have their parents

consent who come to clinic on these days

all of those things that you're going to

include right exclusion criteria then

for that study would be they can't read

and

write they didn't come to clinic on the

same day if they had another condition

maybe maybe I'm looking at people 5 to

10 years old who have acute lymphatic

leukemia but let's say that this patient

also had congenital heart

disease are they also have Down syndrome

am I going to include those then that

may be an exclusion

criteria sometimes there are behavioral

factors that might pose a risk for a

high attrition rate and we want to avoid

those I don't know some kind of study

where you were looking at somebody with

ADHD and they might not be as likely to

continue on the study like you want them

to or might not agree to complete the

questionnaire once they start something

like that so you're looking at anything

that's going to help to control

extraneous variables that might

unintentionally impact your

results sampling in studies qualitative

and

quantitative qualitative you know we're

going to keep comparing these that

purposeful selection it's not random

right we need the people who we think

are going to best be be able to sit down

with us for an hour or two hours and

have a conversation about the

topic maybe it's a study of why people

smoke right so number one you've got to

have somebody who smokes somebody who's

willing to sit down and talk and let you

know what's going

on and for that reason you're looking at

that you're you're looking at a fairly

homogeneous group probably right you're

looking at somebody's very purposefully

chosen because they have the

qualifications that you

need quantitative on the other hand may

use probability sampling or random

sampling you know and that's what I

talked about with our children with

sickle cell no we didn't randomly choose

from the greater population we took

everybody

that would agree every patient that we

had that had CLE cell they were under

the age of six um and they agreed we

signed them up we could have just as

easily said look we have a thousand

patients we're going to choose randomly

500 and then from those 500 we're going

to randomly assign somebody to get

penicillin and somebody to get the

placebo

random essentially every member of the

population when you're doing random

sampling has an equal probability of

being selected so we while we took

everybody that was in that

population once they were on the study

they all had an equal probability of

getting either the the study drug or the

placebo and then every subject selection

is

independent doesn't depend on who just

just we don't say okay well we just took

three that are getting penicillin so our

next one is going to be we don't do that

again it gets thrown in a hat tossed

around pull out a

number this is what I was saying before

a table of random

numbers whoops sorry um a table of

random numbers requires access to a list

of population members systematic random

sampling is a random selection of a

starting point let's say that we give

everybody a number and we just randomly

pick numbers you can close your eyes and

put your finger on a table that's number

79 we'll take 79 close your eyes and

touch it again you're just going to get

a list and you're going to randomly

choose systematic random is a little

more different in that let's say that we

have a list of subjects and we're going

to choose every fifth one so we'll count

1 2 3 4 five we'll take number five 1 2

3 4 5 we'll take number 10 1 2 3 4 5

number 15 and we'll keep going until

we've gone through our list taking every

fifth

subject stratified random sampling is

drawn from stratifications based on some

characteristic that you're interested in

we'll use that to do our sampling

cluster random sampling

is essentially a random selection of

entire

groups and that cluster then we measure

everybody in it everybody in the group

gets measured they're are

cluster ensuring Independence we want to

be sure that every set of data every

person is independent they're

independently chosen

right if somebody is related if they're

related in some way if you have a bias

and you choose certain patients

then you violate that Independence if

you get more than one score well oh this

one I like better so I'm going to take

it versus the first one they did you

violate that at

Independence so pretest and protest is

not independent right you've got you do

a test you do some intervention and then

you test again or time series because

they've already had a

score and an intervention and they're

doing it again you see how that's not

hopefully not

independent has good qualities and

sometimes we want to do that but there

you go the most common thing we do

honestly is convenience

sampling I had to do that to get through

with my dis

dissertation it is where you take

whoever subjects are available in a

convenient way to the researcher I went

to clinic I asked every child that

showed up to Clinic that day if they

would participate it took forever to get

subjects I will say and I really thought

it wouldn't take us long I one time I

decided to go to camp and I got kids who

were coming and checking in at camp with

their parents and I got them to sit down

with me and do my questionnaires really

quickly so that was one way to do it uh

snowball sampling is convenience

sampling too and that maybe I have a

subject and they've done my

questionnaire um um maybe it's somebody

who knows other people if you're looking

at somebody who's homeless they know

other homeless

people maybe they'll recruit them for

you maybe I want to look at cheerleaders

and so I have a subject who's a

cheerleader and she can recruit or he

can recruit other subjects in that meet

those qualifications it decreases your

generalizability because of that but

often times we can get what we need in

that way convenience sampling quite

often is used in a pilot study or a

study to find out if we can even do the

study you kind of test everything out to

see if it works

random assignment then can be done on

top of convenience sampling and I guess

that's kind of what you could say we did

in that we took everybody who came to

Clinic which is

everybody and then from that we randomly

assigned them to the two groups and that

helped with bias we weren't choosing who

got study medicine and who didn't and we

didn't know who got study medicine and

who didn't

so uh hard to reach populations you can

speak in their native tongue if you know

the language you can use experience

sampling methods respondent driven

sampling people who will participate

people who will complete a questionnaire

or send it to you service based sampling

out there is a well is an opportunity to

get hard to reach

areas we can do mixed methods and that's

when maybe we're studying something very

rare rare

disease um or disaster responses we got

a lot of those lately right um so we mix

we may do cluster and referral sampling

where we randomly select a geographic

area that's our cluster right and then

our respondents can identify others who

have that same experience and they'll

refer them to

us sampling size and Power we'll

definitely talk more about these things

but it's important to know that with

qualitative studies we don't go in with

a predetermined idea of how many

subjects we're going to need we're going

to continue to sample we're going to

talk to

people and look at our data begin to

code our data and determine we start to

see Trends and

patterns when we start to see that

redundancy then we start to look at

saturation we think we've got all we're

going to get we've got enough that it's

not going to

change

saturation quantitative samples on the

other hand we're going to need a certain

power tell us how powerful our our uh

data is the greater the number of

subjects the greater the power the

greater the significance and and really

the the magnitude of our results is

going to be much stronger with

quantitative samples if we have a large

sample size and we'll do a Power

analysis to determine how large it ought

to be and then we'll try to get that

number

of subjects

right small samples you're concerned

really that you have a representative

sample and you can be concerned about

type two errors and that is that um

there may be a difference between the

groups but the groups are so small that

we can't pick it

up so when you go to read a subject read

a study and you're looking at the data

that's there look at what's being said

read the study you can't read the title

or the topic the the title or the

abstract and really know what it's about

honestly read the abstract but then read

more information how were the subjects

selected was it

objective was it

random how was the sampling strategy

applied was it consistent throughout or

were there inconsistencies and how they

ran the

study how about how they're assigned to

treatment groups was that random was

that impartial or did the kind of pick

and choosey they wanted to be in the

treatment

group were there enough subjects

included that you can feel like the

results are good and that the

conclusions

are um reliable I guess realistic

true qualitatively you're looking at

criteria

established that were the

characteristics desirable in the

informance did you establish those did

you look for those then did the

informants have those characteristics

did the researchers apply apply enough

effort to find respondents who would be

best for answering informing the

questions and finally that word

saturation that you got enough subjects

and enough information that you weren't

getting anything new that it wasn't

changing what you had that you were

seeing those Trends and themes emerge

and then you've got saturation

using research as evidence for practice

look for threats to validity a lot of

times we talk about validity and

reliability validity population validity

ecological validity is it usable how did

you define the population what were

extraneous variables out there maybe you

caught them maybe you didn't did they

affect the outcome could they affect

outcome

is the setting reasonably similar my

concern for my study was that because I

went to clinic and I and I got subjects

and clinic but then I went to the camp

setting people checking in the

camp I hope they were reasonably similar

my RB decided they were so have they

been replicated if this is a one study

and nothing else gets the same findings

you got to wonder

so in terms of describing adequate

sampling strategies Define you've got to

carefully Define your population you've

got to have inclusion and exclusion

criteria who can participate who cannot

participate that's one way of kind of

controlling those extraneous variables

what's your recruitment plan how are you

going to get subjects to

participate do your power analysis and

determine how many subjects you

need apply the methodology that you

chose and then do everything you can to

maximize retention now in my study I did

a one-time thing and everybody completed

my questionnaire one time I had three

questionnaires everybody completed them

at once so I really didn't have a

retention issue and I told my subjects

and their parents that they could

withdraw from the study at any point up

until they had completed the

questionnaires but once they had

completed my questionnaires they could

not withdraw after

that so for chapter seven you're really

talking about sampling strategies it's

really

important uh if we want to apply our

findings especially to practice right

it's our way to control

bias random selection is always best but

if we can't randomly select from the

population

let's randomly select from our group our

subject

pool um what else survey designs have

additional challenges related to

response rate how many times do you have

a

questionnaire and you look at it you

maybe send it to a thousand people and

10 people

responded that's a terrible response

rate what are you going to do with

that sample size is important sample

size tells us how strong it is and we

want power to be strong it's a concern

only if there's no statistically

significant

results chapter eight talks about

measurement and data collection kind of

more of what we're doing measurement is

a really strong um way to look for and

to collect good evidence measurement

makes it possible to make decisions

right to give us evidence of the

characteristic we're looking at to

determine if it's present how present is

it measurement is based on very

definitive rules for how we measure and

it usually involves numbers some way to

classify or quantify if you will an

attribute we can do that by instruments

or classification a lot of instruments

are out there by instrument it could

mean a tool which is a survey a lyer

scale

survey it could mean a blood pressure

cuff as an instrument right that's going

to give us a number was your blood

pressure higher or lower after the

treatment let's take your blood pressure

heart

ratees things like

that so we need to Define our variables

now we can Define the concept of the the

variable I can Define the concept of

quality of life such as support and

self-concept which is what I did what

does that mean what are the definitions

really of those Concepts but then

operationally I Define those as the

scores the students made on the

instruments used to represent those

Concepts so you really need both when

you're looking at your studies look for

both

and I've said before and I'll say again

make sure you're looking at primary data

it is really the most common data

collection method used systematically

collecting data can be quantitative or

qualitative data you can do it in person

by mail you can do it online but you are

the person collecting the data it's

primary data whoever is the researcher

reporting the results they were the data

collector

that's primary data they're not reading

somebody else's work and then

summarizing it and Reporting on it

because that one step away from the

primary

data could potentially have bias and and

change it just a little bit so we want

primary

data now this is really big to read it's

in your book and I do want you to go

back and look at it but survey design

primary data you can see and secondary

data

all of this is here primary data you're

going to do all of these things

secondary data it's already done you're

just looking at somebody else's data

right primary data what are you going to

do are you going to do a questionnaire

are you going to do an

interview how are you going to do these

things Clos questions give you options

but you have to choose an option it's

close it's

forced and every option is mutually

exclusive of every other option in other

words if I want to know your age I would

say s

to8 I couldn't say 8 to 10 because

Eight's in both right it's not mutually

exclusive so maybe I would say 7 to 8 9

to 10 and then I've got mutually

exclusive and we have our own little

total range of answers dichotomus you

have two answers right yes no true false

it's it's one or the

other scales are used a lot lyer scales

especially there are

continuums of uh where you have a line

and here's one end and here's the other

end and then you would mark yourself on

that line from least to

most when it comes to survey questions

and writing survey questions just

remember they need to be

short they don't need to be ambiguous

they should be unambiguous very clear

they should not be leading they

shouldn't have emotion attached to them

okay survey questions now for the most

part people you wouldn't necessarily

write your own survey

questions you're going to choose a

survey as a researcher hopefully that is

already out there already been tried and

true already has validity and

reliability num so we know it does what

it's supposed to do right if we write

our own if we make our own survey then

we do have to test it for validity and

reliability before we can actually use

it in our study but validity and

reliability of the instruments use

should always be included in the study

you'll be asked in your second paper I

believe about validity and

reliability so look for that information

in your

studies measurement error measurement

error is the difference between what

really is true versus what the score

says is true so we want to know that we

really are measuring what we're supposed

to be measuring

right all right random mirror can be

affected by a lot of different things

we'll talk more about that system IC

area error is just a consistent bias

it's always in there and maybe this is

on purpose maybe somebody just doesn't

know what they're doing but this is why

we ask you to look at peer-reviewed

scholarly Journal articles because those

have been

reviewed by people who know for these

kinds of Errors

so um we want to make sure our equipment

is calibrated if you're using blood

pressure cuff we'll have it calibrated

we'll do blood pressures with it and

three others do they all get the same

result then we can be pretty sure that

it's correct we want to make sure that

it's consistently measuring the way it's

supposed

to reliability tells us how much we can

reproduce this result we want it to be

reproducible right we want somebody else

to take our our information our

instruments our model and methods and do

it and get the same results we got

hopefully if there's stability within

the instrument that's internal

reliability each element measures as

it's supposed to together um you want

stability in that way you also want

stability between Raiders if I am going

to people and asking questions and

rating their responses then I need to

know that but this other person going

and doing the same thing is rating it

the same way that I am and so if you

have more than one person doing

interviews or or rating scales we'll

have everybody sit down together and

we'll do them and make sure we all match

um and then over time you want it to be

able to over time if we test we retest

it should still be reliable to measure

as it's supposed

to the validity validity does it

consistently measure

accurately um it really does reflect the

content it's supposed to reflect uh test

blueprints help us determine content

validity construct validity it does

represent the concepts that we want to

know about uh we want concurrent

predictive discriminate Criterion

related validity lots of different ways

to check um and talk about validity but

the bottom line reliability and validity

it measures what it's supposed to

measure and it can do it consistently

over

time diagnostic

measures

sensitivity of an instrument says that

it can detect

disease

specificity it can detect when the

disease is not there and responsiveness

can it test or can it check can it tell

uh when an intervention is affected does

it pick up that

change collecting data using instruments

a lot of instrumentation is online now

it's easy right it's low cost I can send

you a survey monkey and have you do it

it doesn't cost anything um it's

accurate

right it's convenient for the

participants and for the researcher you

just can't interact with people

sometimes I think if you talk to people

uh maybe they have a better feel for a

desire to help a desire to want to do it

but you do have that risk for systematic

sampling error too in terms of online

social media you can get a large

geographic area in a study it is good

good for hard to- reach

populations and it's still relatively

low cost right I saw something on

Facebook a few days ago a survey that a

nurse was asking people to

do um disadvantages privacy right

ethical concerns potentially because you

may be seeing information shared by

others other people may be seeing who

agrees to

participate right and then self-

selection error may be a RIS maybe they

said oh yeah I'm going to sign up for

that and they really don't have the

disease that you're testing for or

looking

at Technology based pretty much you get

higher response rates it's convenient

for everybody there may be an expense

though do I have to set up computers or

tablets um for people to have access to

to do this I can go to the mall and I

can have tablets but I have to buy it's

a lot of equipment

a lot of variation in the device and how

they work a lot of them are very user

friendly and a lot of devices are not

and it depends a lot on the person an

older person may have trouble with

technology interviews are really good

you can sit down and get a lot of

information you can talk to the person

and you can begin like with qualitative

you begin to hear those

themes and that information is personal

but it's time consume me to sit down and

have an hour an hour conversation with

somebody when I could give them a

questionnaire right and that training

that you need to do an interview we need

to know we're asking the same questions

we're looking for the same information

you're listening for some things that

that should be there so you do have to

train to do those

interviews focus groups um have

benefits um good information with a

group group of people they may um not

want to share with each other U it's

expensive can be time consuming you need

a place to meet and all those people you

probably need some coffee and things

like that so it also requires a skilled

facilitator many times have you been in

the classroom um and seen that it just

kind of got out of control suddenly and

so focus groups can do that if

somebody's needs to be trained to pull

the group back in and stay on topic it's

really hard to

do observation is another one it's not

invasive um it gives you a lot of good

information and we talk about all the

time how much we get from patients just

observing them and so there you go it

has disadvantages it can be timec

consuming depending on how much

observation is needed and if somebody

knows you're watching them they may

change their behavior so there you go

and you can have direct observation and

indirect observation versus something

may be recorded or something like that

but again I think you've got a little

bit of change noted if you know you're

being recorded how does that

impact uh secondary data that's where

you have data collected by somebody else

maybe even for another purpose but now

you have access to this big huge Data

Bank and I think I'm going to go pull

and see how many people in the US were

admitted for cardiac conditions between

the ages of 40 and 50 in 2023 you can

there's a database that you can go and

pull that information for free um and so

there that's secondary I didn't go

collect that data but I know where to

get it and I went and looked at it's

very efficient very convenient it

greatly reduces the time you need to to

study to go and collect all that data

yourself right but you do have

disadvantages in that sometimes the data

is not complete maybe it's not accurate

how do you know those blood pressure

costs were were checked for

accuracy and you really can't control

how that data was collected and it is

retrospective in a bit it's already

happened and you're looking back on that

data big data is just that looking at

vast amounts of data but using data

science method somehow you can

do really

large what's the right word really large

evaluations of data um but you have low

error rates it's it's got a high value

and it's really has a role in

evidence-based nursing right now we do

some of those things in terms of

systematic reviews but this is even on a

grander scale than that

if you're doing data collection you do

need a procedure for data collection you

would need to write a code book uh

things as simple as male is one female

is two and on and on um maybe you would

code

ages and so all of that becomes

important

right how are you going to manage the

DAT how are you going to organize data

how are you going to store the data

everything has to be locked up I

panicked I had mine locked in a locked

box to transport it had a friend who had

a wreck she had a file box that had a

wreck and there was her data spelled on

the highway so you violated

confidentiality got to be careful mine

is in a locked box and I still have it

from 20 years ago by the way because I

don't know what to do with it I guess I

should go shred

it can't share it that individual data

and we make a point of telling our

subjects that the data is not going to

be identified individually but it's

going to be reported as group data and

that's what we have to

do so really measures and data

collection are a way of doing things

systematically the more systematic and

detailed you can be the the higher your

uh results the greater the quality of

what you're doing is look for an

existing instrument remember we talked

about validity and reliability if you

make an instrument you've got to

establish those things first you'll have

to Pilot it find an instrument that's

already out there that that's already

been done on we already know it measures

what it's supposed to measure we already

know it does it consistently over time

so and then be feasible you want to do

something

that's you've got the time to do do you

got the resources to do you've got the

finances to do you've got the people to

do it and you've got the time to get it

done your delivery method needs to be

consistent with what you want to do in

your study and your data collection

protocols have to be

defined so in this case we looked at

measurement quantifying

characteristics lots of different data

can be Quant Quantified and measured

measurement error is always present but

we do everything we can to make it

minimal instruments have to be valid and

reliable if you're writing your own

instrument you've got to establish that

if you're using one that's already been

established you've got to report

validity and reliability you're going to

be asked to do that in your week five

project and then you've got to collect

data systematically and I believe that's

it for the week isn't it guys

yay thank you for listening