aim 1 data analysis Flashcards
Define inductive and deductive coding. Example
Deductive: coding to test whether data are consistent with prior assumptions, theories, hypotheses.
Specific questions for site directors about qualities of OS that increase credibility. Pros/cons of individuals having criminal backgrounds- how that can perhaps help or hurt credibility
Inductive: use data to develop concepts/themes. Emerge from the data instead of pre-determined
Drug market dynamics: don’t have pre-conceived assumptions about what types of drug market dynamics are associated w more or less violence or more or less difficult to intervene on.
Define axial and open coding. Example
Open coding: first step in developing a set of codes to describe and classify the qualitative data. Phrases or words that are salient, concepts related to my SEM, things that are repeated within an interview.
Ask about police activity, based on answer may create codes for: positive police interaction, disassociation with police, collaboration with police
Axial coding: step after coding, relating data together within and across participants. What are the connections among codes? Creating hierarchy of codes.
Looking across sites or within site OS talking about police, may see that people often mention specific officers and that relationships with specific officers are important or fragile or..
Define cross case comparison Examples
Cross-case comparison: Previous quantitative analyses, qualitative analyses from interviews, and mixed analyses from program implementation data will all be integrated to create a holistic description of each case. Cases will be compared and both consistent and inconsistent findings will be explored.
Relevant groupings that emerge and whether that is associated with overall trends of effectiveness or ineffectiveness.
Notice overlap with pattern of quick effectiveness followed by declined or plateaued effectiveness associated with staff turnover or staff loss that were not rehired, as compared to sites who don’t talk about staffing issues who have more consistent results
define joint display and example
To facilitate cross-case comparison, joint displays will be created that may for example, overlay the ASCM findings of SSB effectiveness overtime with key timepoints at the site (e.g. changes in site leadership, changes in underlying patterns of violence or drug markets).
Visual display that includes quant and qual data
ASCM graph findings with key moments in site timeline
map of neighborhood w quotes
meta inference definition and example
The final stage of data analysis will be the interpretation phase in which all data sources and analyses will be integrated. This integration will result in meta-inferences, or inferences that could not have been achieved by looking at the quantitative or qualitative data separately
Quant- how “effective” the site, qual- what makes a site effective. Powerful individually but more powerful to integrate those two strands to show for example change in police leadership associated with jolted trends across several sites
define data transformation and give example
It is expected that program implementation data will include both qualitative and quantitative data and will not be provided uniformly across sites. Qualitative data from both program implementation data and semi-structured interviews may be converted into quantitative data and entered in an excel spreadsheet to facilitate cross-case comparison.
Interviews ask about for ex. Mental health support for staff, that turned into a binary variable and used as part of a description of benefits
What is intercoder agreement/reliability? Why 80% intercoder agreement for reliability? Talk about cohen kappa
Probability that two or more coders code the same text with the same code
Line up codes of the two reviewers and calculate the percentage of agreement- can do on a subset of themes or full set of themes, subset of transcripts
Agreed/agreed+disagreed
Acknowledging the role of researcher identity in coding and qual data analysis, not expecting or looking for 100% agreement, but benchmarks are .8 is a standard for strong agreement or high reliability, .7-.79 adequate.
