seeks to find meaningful relationships between variables

• statistical analysis via
  • retrospective or prospective studies
  • controlled studies
  • observational studies
• descriptive statistics
  • presentation, organization, and summarization of data
• inferential statistics
  • ID relationships that can be generalized from our study population to the larger population it represents

• collect observations about subject/topic of interest
  • ex: case control studies, survey studies, HMIS data

• data is collected, change is made, data is collected again following change
• must have control group

• correlation- two events (A,B) observed to happen together
• causation- A is the cause of B
• statistical analysis tells you that A occurs in correlation with B (therefore A causes B is FALSE)
• statistical correlation alone never proves causation
• gold standard for determining causation is carefully designed, controlled experiments

• strength- numerical strength of correlation
• consistency- observed in many places at many times by many different observers in different circumstances
• specificity- effect limited to certain observations in certain specific situations
• temporality- A must occur before B
• biological gradient (dose response relation)
• plausability- scientific credibility of relationship
• coherence- possibility of causal relationship should not conflict with what is known about natural history and biology of situation
• experimental evidence- if you take away intevention, does "B" does happen
• analogy- reason from similar phenomena

• statistical- we are 95% sure...
• economic significance
  • explanatory variable has meaningful and plausible influence on dependent variable (will it make a significance public health difference?)

statistical evidence is necessary but not sufficient for economic significance

• before you can analyze data, first need to collect with an understanding of what you are collecting
• not understanding the data you have collected, or using the wron data, can lead to incorrect conclusions and drive misguided changes in policy
• big part of what determines whether data is good is survey design and sampling

• systematic observation- or measurement of various features of behaviors in the world
• logical explanation- in the form of a theory or model that makes sense according to basic rules of logic and accepted facts
• prediction- in the form of hypothesis, based on theory, of what we will observe if the theory is true
• openness- meaning of methods used to produce evidence are clearly documented and made available for review (this allows for replication- repeatin study to see if results hold)
• skepticism- researchers scrutinize and critique each other's work, a process reviewed to as peer review, in search of possible shortcomings or alternative explanations

• you are interested to answers in the survey, not intrinsically but because of their relationship to something they are suppose to measure (ex: failure to use a condom = risky sexual behavior)
• questions are only good if they are:
  • reliable- providing consistent measures in comparable situations
  • valid- answers correspond what they are intended to measure
• measurement refers to the process of systematically observing some feature of characteristic of the world and then recording it

• start with a construct
• find a way to measure it
• error is a part of that measurement
  • ALWAYS have some measurement error in studies

• first step is figure out what you want to measure
• must be defined carefully and precisely
  • some concepts are not easily defined such as poverty
  • require value judgements
  • manifest vs. latent constructs
  • dimensions
• once you have the concept, it is time to operationalize
  • ex: what did the US do with poverty?

• providing consistent measures in comparable situations
  • when two respondants are in the same situation they should answer the same question in the same way
  • any difference in answers stem from actual differences
• good questions
  • researcher's side of the question and answer process is entirely scripted so that the questions as written fully prepare a respondent to answer questions
  • questions mean the same thing to every respondent
  • kinds of answers that constitute an appropriate response to the question are communicated consistently to all respondents
  • avoid inadequate wording
  • ensure consistent meaning of all respondents (can do this via pre-testing)
    • need common frame of reference