L7 - L8

Question 1

inferential stats

Answer 1

techn tht help in generalizing from a sample to a larger group using stat tests of significane
-based on assumptions of what is true for a sample and whats true for larger pop (generalize)

Question 2

causality

Answer 2

many policy issues address q such as “if we were to change a would b get better?”

reverse:
when what is believed to be the cause is actually the effect (determine this by time sequence of the 2 varibles; which occured 1st?)

-effect y occurs whenever cause x occurs
-effect y never occurs without cause x having occured already
-effect y occurs when cause x hasnt occured + always occurs when x has ocured

Question 3

associations

Answer 3

causality and correlations
-dont necessarily mean eachother
-association can be due to causality
-confounding: if 2 are associated bc of mutual association with another variable
-spurious: when 2 v are correlated but have no causal relations
-causal relations: if apparent cause (ind v/predictor) is changed the hypothesized effect (outco/dep) will change in response

Question 4

univariate analysis

Answer 4

type of data which consists of observations on only a single charact/attrib
- 1 v
-explores each v in data set sepearately
-range of values and their central tendency
-describes distrib pattern of v

Question 5

bivariate analysis

Answer 5

involving/depending on 2 variables
-based on how 2 v simultaneously change eachother
-covariance: the idea tht 2 v vary together; knowing the value of one can lead to calculating the value of the other
-scatter plot
-correlation coefficient
-how they associate and simultaneously change together

Question 6

stat significance

Answer 6

-formal way of assessing whther observed associations are likely to be explained by chance alone
>0.10 not significant
<_0.10 marginally signi.
<0.05 significant
<.01 highly significant
-significance testing: used to help make a judgement about a claim by addressing the q; can the observed dif be attrib to chance?

Question 7

correlation

Answer 7

-quantifies the extent to which 2 quaniti v “go together”
-high values of x and y = pos corr (opp/low= neg corr)
varies betw -1 (neg) and +1 (pos); 0 = no relation
-correlation coefficient: stats that quantify the relationship betw 2 variables (betw -1 and +1; denoted r;no units; sign determins if its neg or pos; magnitude determines correlation strength)
-cc of determination: get from squaring value of r (r2) to get proportion of variance in 1 variable shared by the other (quantifies the proportion of the variance of 1 v explained by the other)
-regression and correlation assume v are interval/ratio and relations are linear

Question 8

types of error

Answer 8

-survey may be designed to measure abstract conceepts (ie vaccine hesitancy)
-error is dif betwn obtained values and true values (usually for large pop of interest)

Question 9

error: sampling error

Answer 9

-due to respondent selection
-when a sample of the pop rather than the entire pop is surveyed
-prob sampling helps as its in a mathematical form

Question 10

error: coverage

Answer 10

when the list from which the sample is taken (sampling frame) doesnt correspond with the pop interest

Question 11

error: non response

Answer 11

item: respondent participates but skips some q
unit: designated respondent doesnt participate in study (limits representativeness)

Question 12

errors: measurement

Answer 12

when the measure obtained isnt an accurate measure of what was to be measured
-could be due to respondent (accurate ans? q wording and sequencing)
-or interviewer (implications can lead to innacuratecies)

Question 13

stat impact of error

Answer 13

the goal of identifying error isnt to elimnate but to minimize them (acces their stat impact; affects analysis of survey data)

total survey error approach: based on anayzing the several dif sources of error in surveys and considering how to minimize them in the context of practical constraints

Question 14

random error

Answer 14

-not systematic
-if ppl dont think hard enough about a q to ans it correctly they may get error
-no systematic tendency in either direction = mean of 0

Question 15

More issues

Answer 15

respondent selection issues
-sampling, coverage and non response error (unit)

response accuracy issues
-non response error (item), measurem error (due to respondent/interviewer)

survey administration issues
-post survey error, mode effects, compatability effects

Question 16

equations

Answer 16

straight line:
Yi=bo +b1Xi+Ei (etc 1…2…3…)

bi - regression coefficient for predictors (gradient/strength of relationship)
bo -intercept (value of y when x = 0)
-ordinate (where regression line crosses the y (vertical) axis

Beta values:
b1 - regression coefficient for v1
b2 - for v 2
bn - for v nth

Line of best fit:
Y=a+bx+e
y - dependent
b - independent
e - error
value of the dep v at any value of the ind v can be determined

Question 17

sci surveys

Answer 17

• samples memb of the defined pop in a way such tht each memb has a known prob of selection (uses procedures drawn from accepted stat theory and uses them on account for the pop not covered by sampling design)
  -collect data from a sufficient numb of sampled units to be generalized to the pop
  -coverage: sample memb of the defined pop in a way that each memb has a known non zero prob of selection (no sampling bias when attempting to generalize)
  -stages: decide objectives (re), determine target pop choose survey mode and design, choose sampling frame and method, write ? (link to op def), pretest questionare and recruit respondents, ask questions, process data, analyze results

Question 18

ethics

Answer 18

the use of monetary incentives in a survey could be considered as kindn tht envokes a norm of reciprocity
-adhere to re design, protocal, and informed consent etc (and analytical plan) all good for data integrity

Question 19

mode effects

Answer 19

-results from self admin and interviewer admin surveys are sometimes dif (=mode effect, a dif in responses to survey question attrib to the mode in which the ? is admin)
-sensitivity of ? may be greater for some than others
-social desirability bias: a type of response bias that is the tendency of survey respondents to answer questions in a manner that will be viewed favorably by others.
-recency effect: ppl mainly remember the last response opt read by interviwer and may be favored by respondents (less likely for self admin survey)
-a complicated ? with many opt may be dif to comprehend