mini quiz 1

reliability is

true score/ observed score.

=

true score/ (true score + error score)

reliability of the observed score becomes higher if error is reduced

observed score is true score + error score

conceptualizatioin

the process of specifying what we mean by a term

• what do u mean by “healthier”

operationalization + operational definitions

the process of connecting concepts to observations

specifying exactly what will be observed and how it will be done

definition of true score and error score

true score= perfect refelction of true value (theoretical score but can never truly be known)

error score = diff between observed and true score

sources of error and what r they a major concern for

random error (due to chance)- major concern for reliability

systemic error (error of measurement)- major concern for validity

reliability means it is consistent . 4 ways to test consistency of measurement tool

1. test- retest

2. parallel forms

3. inter-rater/intra-rater

4. internal consistency

test retest reliability

evaluates stability over time

• give same test to same ppl at two time points and see if same results obtained

concerns: need to be sure charecteristic measured does not change over time (eg. food eaten changes per day) and be aware of testing-effects

parallel/alternate forms

measjre of equivalence

• give two diff forms of the test to the same grp and compare results . often want to eliminate “practice effects”

• eg. website vs paper

rater reliability

measure of agreement

inter-rater: consistency between raters

• have two raters judge same event/behavioiur and assess agreement

intra-rater: stability of measure by the same person (to assess effects of bias, fatigue etc)

internal consistency and how to measure

measure of how consistently each item measure sthe same underlying construct

• do subscales speak to similar underluying construct?

• eg. 100 item questionnaire of food attitudes

• eg. contradicting responses

correlate performance of each item with overall perfomrance across participants

how to measure

split-half reliability: randomly divide items into 2 subsets and examine consistency in total scores across two subsets

cronbachs alpha: avg consistency across all possible split-half reliabilities. 0= not reliable. 1= perfect correlattion. questtionaire shd be trustsed

validity measures what it shd

truthfulness, accuracy and authenticity

we cannot have total validity without reliability

4 ways to assess validity

1. face

2. content

3. criterioin

4. construct

face validity

how well the instrument appears (at face value) to test what it is supposed to

• eg. “does ts look like it will be a good test of ones nutrition knowledge?”

content validity

how well items represent all relevent items

• ask an expert. does this instrument measure everything its supposed to?

criterion validity

gold standard

ability of the tool to predict results obtained on an external criterion or reference standard

• criterion shd be a valid indicator of cariable of interest and relevent to variable being measured

• eg. FFQ (target tool) compared with objective measure such as blood biomarker (criterion)

parallel/alternate forms reliability vs. criterion validity

reliability: give two diff forms of same test to same grp

validity: compare results from two diff tests on same grp. comparision to gold standard known in the field (eg. blood or urine biomakres)

construct validity

extent to which test results are related to underlying abstract construct

• basically using many factors and comparing it to ur expectation. if it chekcs out, then it has construct validity

2 types of construct validity. convergent and discriminant

convergent validity: two measures of a siimilat construct shd be correlated.

discriminant validity: two unrelated measures shd not be correlated. eg. tool that measures healthy eating and a tool that measures financial knowledge should give different scores.

absolute validity requires

a true gold standard that is an exact measure of what it si intended to capture. for nutrition studies we seldom achieve abs validity

reliability= consistency, validity=correctness

statistics terms- parameter

measure of a population charecteristic

statistic

measure of a sample charecteristic

descriptive statistics

summarizes data and provides basic info abt the distribution of scores in a samepl

inferential statistics

inferences from the sample to a larger population

bar plot and pie chart display

categorical data

histogram and box plot display

numerical data

what are and how to calculate mean, median and mode

mean= arithmetic avg

• sum of scores/number of scores

• miu for populations, x bar for samples

median= midpoint of distribution

• half of scores above and half of scores below

mode= most freq occuring score

• useful for categorical data

normal vs positive vs negative skew

normal curve: mean=median

positive(right) skew: mean>median (mean is pulled to right). tail is to the right

neg skew: mean<median, mean pulled to the left

left skewed means

if tail is left. it pulls mean to be smaller than median

does central tendency tell us abt the spread or variability of the data

no

definiiton of variability

the degree of spread or dispersion in a set of scores

range

diff between highest and lowest scroe

SD

avg diff of each score from mean

interquartile range

diff between first and third quartiles

measures of position- percentiles

value below a certain percent of the data lie

• eg. 25th percentile is 25% of the data lie below this number

quartiles

set of numers that divide the data into 4 equal parts

what to use for categorical data, normally distributed data, skewed data?

categorical: proportions/ frequencies

normally distrib: mean and SD

skewed: median and IQR or 5th to 95th percentile

what are independent variables

txs or conditions under control of the researcher. eg. age

dependent variables

the ourcome of the resarch study, sensitive to changes in the independent variable

inferential stats

when we find a diff between grp means in our sample

• want to know if it represents a real diff between target grps or if its just by chance

null hypothesis

hypothsis that predicts no effect, or relationship between independent and dependant variable

• the null is assumed to be true in the absense of other info

• can either reject or fail to reject the null hypothesis

alpha= probability of makingn a type 1 error (false positive)

often set at alpha= 0.05

• 5% chance that we will reject the null when it is actually true

• the alpha is determined before using statistical tests and used to determine if a result is satistically “significant”

beta= probability of making type 2 error (false negative)

we incorrectly fail to reject the null when there IS a diff between grousp

beta usually set at 0.20

20% chance of failing to reject null hypothesisnwhen there is acc a diff between grps

t-test is

we use the t value to calculate a p-value

ratoio of mean to SD

requires:

1. mean for each grp

2. sd/variance for each group

3. sample ssize (degrees of freedom)

the greater the variability in distributions, the more likely

the diff betweeen the meeans is due to chnace

t value is

difference between grp means/ variability within grps

t test for independent means vs dependent means (paired t test)

independent: used when 2 independent unrelated grps are compared. eg. veg intake between male and female studnets

dependent: when grps are related. eg. diff in veg intake in the same individuals on monday vs saturday