Grounded Theory for Qualitative Research

Glaser and Strauss (1967); Strauss and Corbin (1990); Glaser (1992)

based on Dick, Bob (2005)  Grounded theory: a thumbnail sketch

...

Grounded theory begins with a research situation, and observations regarding such a situation. Constant comparison is the heart of the process. Compare interview data with other interview data, then make theories, compare that theory to interview data, etc. These comparisons are often noted in margins of text, and encoded to the likes of the researcher. Relative structure of themes and how themes relate to each other can be written as memos, and then memos are sorted as a skeleton upon which writing can occur. 

gtphases
Grounded Theory, process

 

“In short, data collection, note-taking, coding and memoing occur simultaneously from the beginning. […] The theory is emergent — discovered in the data, Glaser will say.”

What most differentiates grounded theory from much other research is that it is explicitly emergent.  In this respect it is like action research: the aim is to understand the research situation.  “The aim, as Glaser in particular states it, is to discover the theory implicit in the data.”

 

Questions to judge grounded theory research: Does the theory fit the situation?  Does it help the people in the situation to make sense of their experience?

engine
convergent interviewing (see Note-Taking)
  1. Data Collection
    • Observations
    • Interviews
  2. Note-Taking
    • Glaser recommends not taking notes or recording. This stimulates natural rapport and leaves you with a summary rather than being bogged down in word-by-word transcriptions.
    • Dick recommends recording the session, taking very minimal notes, so that you can go back and verify your effectiveness.
    • Convergent interviewing can also be used among a group of researchers. In this, interviews are found touch on similar subjects, and similarities and discrepancies are identified for further study and appropriate theorizing:
  3.  Coding
    • Start by asking general questions about the situation and how the participants manage their situations.
    • On the second interview develop a code in the margins that compares the first interview to the second interview.
    • When repetition of codes begins to emerge, convert them into memos.
    • Memos and categories of memos can be organized in various sub-groups in a hierarchical or network fashion.
    • The process of interviewing stops upon saturation, where no new information in a category is added.
  4.  Sampling
    • The sample of interviews should be emergent, in that, when new categories emerge, differences in those categories should be sought out to study.
    • You shouldn’t limit your research to the initial sample if it ends up not being completely representative of the diversity of your population.
  5.  Memoing
    • “A memo is a note to yourself about some hypothesis you have about a category or property, and particularly about relationships between categories.”
  6. Sorting & Writing
    • Reorganize the memos as needed to describe the phenomena/situation to an outside audience. This guides the writing structure.
  7. Using Literature
    • It is not apparent at first what literature will be appropriate to cite in writing until the observations and interview processes begin.
    • The literature to be cited should be treated as equal to the data you’ve collected.
    • Glaser warns about reading too specific of literature before interviews, in that the coding and memoing could guide you to make assumptions that won’t exist during your observations/interviews.
    • If an apparent disagreement between your emerging theory and the literature exists, don’t assume that your theory must be wrong.  You should seek to extend the theory so that it makes sense of both the data from your study and the data from the literature.

“In short, in using grounded theory methodology you assume that the theory is concealed in your data for you to discover.  Coding makes visible some of its components.  Memoing adds the relationships which link the categories to each other.”

Advertisements

Glasses: This Object and its Origin

Objects exist in multiplicity.

The physicality of this specific pair of glasses is just one perspective in defining these glasses.

IMG_4674.png
these are my glasses

From where did these glasses come? Of what materials are these glasses crafted? What parts of these glasses are unique to myself? What important milestones have these glasses passed? Where and when do these glasses get used (or not used)? How do these glasses effect my life (physically, emotionally, etc)?

From where did these glasses come?

These glasses were ordered on Zenni Optical‘s website.

Zenni Optical began in 2003 in the San Francisco Bay Area. They boast more than 6,000 styles of frames online, including men’s/women’s/kid’s frames and lenses, as well as sunglasses and sportswear glasses. Zenni owns at 248,000 sq. foot facility that “houses state-of-the-art Rx and Edging Labs.” While the pictures show the San Francisco offices, the manufacturing plant is located in China (according to the Terms of Use), and most US orders are made in China, shipped to San Francisco, then to your desired location.

Of what materials are these glasses crafted?

These frames are specified as Browline Sunglasses #732021

browline_glasses.png
the specifications for my frames

Browline Sunglasses #732021 are made from a mixture of acetate (Cellulose acetate) and silver alloy full rim and silicone nose pieces. The lenses are made of 1.61 High Index plastic,  though no specific plastics are specified over 1.58 refractive indexes. There is an anti-reflective coating on them to reduce glare that is of unknown/non-specific composition.

What parts of these glasses are unique to myself?

The prescription: I have myopia, or nearsightedness, in that I can’t see at distances well, but reading-length vision is appropriate–indicated by the negative sphere (SPH) lens power. (I also have a slight myopic astigmatism in my right eye).

prescription

PD stands for pupillary distance, or the distance between the pupils. According to an infographic on Zenni Optical’s website, adult PDs usually fall between 54mm-74mm.

The lenses are often dirty, as I only clean them with cotton part of a shirt I am currently wearing. The frames are not level with the table, as I do not adjust them to uniquely fit my ear and nose-bridge shape.

What important milestones have these glasses passed?

I broke my previous pair of glasses on July 10th, 2015 at approximately 1:15am. I took them off while sitting on a front porch in the Short North of Columbus, Ohio, and a friend stepped on them.

I ordered these glasses on Dec 12, 2015. They arrived Dec 21, 2015. 12 of 15 selfies I’ve taken since Dec 21, 2015 have been while wearing these glasses. I’ve now owned them for 56 days.

Where and when do these glasses get used (or not used)? 

I wear these glasses most days. I also have contact lenses which I wear on occasion. I normally only wear contact during sports with physical interaction (like volleyball) or when I plan on running long distances (sweating a lot).

When I put on my glasses in the morning, I grab them from the top (or second to top) shelf on my bedside table. I usually set them next to my iPhone while it is charging, a glass of water, and my keys and wallet. I wear them while commuting and while in class. I often wear them while at the gym but find myself taking them off when doing some ab exercises on the ground. I take them off to shower at home, and I place them next to the hand soap dispenser on the left side of the sink.

How do these glasses effect my life (physically, emotionally, etc)?

Physically these glasses are a hassle when wearing them in the rain. I commute using public transit and I’ve only worn them once in the rain. They also leave a slight impression on the bridge of my nose from wearing them on a daily basis.

IMG_4779e.png
impression of the silicone nose pad after daily wear

I originally bought the glasses because of the 1960s academic style, epitomized by Henry Crane (portrayed by Rich Sommer) in AMC’s MadMen. I think of glasses as an essential part of my personal style. This particular style emphasizes a brow line that I personally think is weak compared to other’s brow lines.

25
Rich Sommer as Henry Crane in AMC’s MadMen, styled by Janie Bryant

 

Research Methods: Readings II

“Why Should Engineers and Scientists Be Worried About Color?” Link
Bernice E. Rogowitz and Lloyd A. Treinish

figure1
the same data but mapped with mathematically equivalent scales, but one adds a threshold of significance (sea-level)

With the color map on the left, elevation is a continuous variable. However, its corresponding color scale ranges among approximately 5 discrete color categories. This confuses the reader.

“One result of this work has been a set of colormaps which take into account the data type, the spatial frequency of the data, and properties of the human perceptual system.  These colormaps are all designed to create more faithful impressions of the structure in the data.”

Data types include:

  • Nominal (no mathematical relationship)
  • Ordinal (occur in an order, but no mathematical relationship)
  • Interval (experimentally determined relationships)
  • Ratio (equal measurement between values, with zero included)

How do we choose color for these data types, so that our perception of color judgement matches the comparison of the data types?

figure3
Hue, Saturation, and Luminance: three perceivable dimensions of color

Hue, by itself, is not known for producing accurate judgments of a coded variable with varying magnitude. How to choose between a saturation-based color scale or luminance-based one? If there are great frequency shifts, use saturation-based color scale to see the graduation of changes. If there are small frequency shifts, use luminance-based color scale to emphasize distinctness in extremes.

Two special cases for color ranges: Segmentation and Highlighting

  • “In segmentation, the analyst’s goal is to look at the whole range of data, but partitioned.  If the segments are derived from interval or ratio data, it is important to preserve the perception of order, that is, that the order of the segments matches the order of the data values.”
  • “In highlighting, the analyst’s goal is to focus on a limited range in a variable and study how this range expresses itself in the data set.  The analyst, for example, may want to probe the exact ranges where the dose of a radiological treatment affects distant healthy tissue, or the particular magnitude at which the wind changes direction in a meteorological simulation.”
figure7
four color maps of photochemical pollution levels with a rainbow-scale (top-left), a isomorphic color-scale (top-right), segmented color-scale (bottom-left), and a highlighting color-scale (bottom-right)

So why the three other color scales? Firstly, an isomorphic color mapping has equal perceptual changes in color between equal intervals of the data. This is the “most” truthful.

Let’s say an alternate analysis is needed (maybe even quickly), a segmented map could show dangerous areas. In the bottom-left map, maybe 140 and above are toxic levels of chemicals. They are easily picked out of the map. What about a specific area, like low levels of a chemical that needs to be evenly spread throughout an area? The lower-right map shows off areas lower than 50 with perceptual ease. The higher areas, sufficient in chemical concentration, are not distinguished between.

 

“Visual Math Gone Wrong” Link
Robert Kosara

metro-change
from the US Census Data Visualization Gallery

So what’s going on here? We have some population arithmetic. Important statistics, but what’s going on visually? Area arithmetic: something perceptually is four times as hard to grasp as single dimensional arithmetic. (And outlines for negative population double coded with the subtraction symbol. A good effort, fill would have been better without an outline, maybe on a toned background. But just use a bar chart since population is only a single dimension.)

Where do bar chart comparisons fail? Only once you get to vastly different scales (near 50x-100x difference between minimum and maximum values). Otherwise, comparisons among column components as well as cross-column comparisons are possible.

 

Ch 6: Visualizing for the Mind from The Functional Art
Alberto Cairo

“The ability to anticipate what the brain wants to do can greatly improve your information graphics and visualizations.” —PREATTENTIVE FEATURES

The detection of object boundaries is based off of variations of light intensity and color, and on how well the edges of the things you see are defined.

The brain is much better at quickly detecting variations in shade than in shape.

variations_in_color.png
notice how color effects the visibility and ease-of-search when iconography is plentiful and too similar

Gestalt School of Thought preaches that the brain can recognize and sort differences in patterns because conscious thought can catch up with it. Examples include proximity, similarity, connectedness, and closure.

Cleveland and McGill (at Bell Labs 1984) published a groundbreaking study that pitted visualization methods against data perception accuracy. Most notably, position along common scale (single dimension with same starting point) and position along non-aligned scale (single dimension with same scale but different baselines) allowed accurate judgements. However, estimations in color saturation, color shading, and curvature were the least accurate.

cleveland_mcgill_perceive.png
10 vs 7 comparisons in terms of single-dimension, circular area, and hue

Stereoscopic depth perception (the difference between left eye and right eye images) allows humans to see in 3D however, there are monocular cues that also assist in 3D vision, including saccades, shadows, relative size, and detail/horizon blur.

“Color Use Guidelines for Mapping and Visualization”
Cynthia A. Brewer

schematic_color_legend

One-Variable Color Schemes

Qualitative Schemes: categorical, no computational relationship between states

  • when using color between categorical data, distinguish them with differences in hue, and slight differences in lightness… not equal lightness. (Why? the physiological system to distinguish hue has poor shape- and edge-detection, the lightness difference will clarify boundaries.)
  • the more categories (the closer your hues), the greater the difference in lightness needs to be. also, consider why you’re displaying that many categories.
  • if any area is small by comparison to other areas its spatially proximal, a greater contrast in lightness would benefit detection.

Binary Schemes: on- or off-states, yes- or no-states

  • differences in hue and/or lightness can be used equally effectively.

Sequential Schemes: ordered, low- to high-values

  • mapping low values and high values depend on the display. treat the display or background default as low, and the addition of ink or light (etc) as an increase.
  • pure black-and-white schemes will have a disadvantage when it comes to the default of the display (areas with no data, like water on a country map) will appear to be on the sequential scale as a zero point.
  • it is not recommended to vary a sequential scheme by saturation only when there are more than 3 sequenced points in your range
  • if using more than one color in a sequential scheme, vary lightness and darkness singularly and evenly, despite hue.
  • full-spectrum schemes (rainbows) are perceptually disadvantageous because yellow is perceptually lighter in hue, and darkened yellow is perceptually desaturated, leading to misperceiving information where, often, none exists.

Diverging Schemes: ordered, with a noteworthy midpoint (or similarly critical point)

  • recommended to use two hues and darken them as their absolute value moves away from the critical point.

Two-Variable Color Schemes

Qualitative vs Binary Schemes

  • choose two hues (one for each binary) and vary lightness by qualitative values
  • increased saturation on the binary value you wish to emphasize

Qualitative vs. Sequential Schemes

  • choose three hues for the qualitative, and vary lightness by sequential values

Sequential vs. Sequential Schemes

  • a logical mixture of two hues and a lightness. lightest represents low sequence in both variables; increased hue and lightness in each color paired with each sequential schemes; increased hue combines in both to show high sequence in both variables.

Balance Schemes

  • neither end of a balance scheme should be emphasized, choose hues carefully to match saturation as best as possible, and vary hue equally between the domain.
  • a special case of sequential/sequential scheme, but one variable increases as the other decreases and vice-versa: they cannot diverge from this formula.

Diverging vs. Diverging Schemescolor_schemes_diverge.png

 

Research Methods: Readings

The Eyes Have It: A Task by Data Type Taxonomy for Information Visualizations
Ben Shneiderman (1996)

  • Information Seeking Mantra: Overview first, zoom and filter, then details-on-demand
  • An issue arise as information systems grow in size–how to design for two cases: known-item search and browsing for patterns
  • (In 1996) GUIs are becoming more advanced, more detailed, and more colorful
  • Scientific visualization has the ability to display, explain, and make comprehensible complex phenomena. Abstract visualization has the ability to spot patterns, outliers, or gaps.
  • The bandwidth of visual information consumption is higher than any other sense.
  • Task Taxonomy:
    1. Overview
    2. Zoom, on subset of interest
    3. Filter, hide uninteresting items
    4. Details-on-Demand, when requested or needed
    5. Relate, show relationships
    6. History, to support undo and replay and refinement
    7. Extract, sub-collections and query parameters
  • Data Types
    1. 1-dimensional (like alphabetized, text)
    2. 2-dimensional
    3. 3-dimensional
    4. Temporal (like one-dimension but with ability to overlap)
    5. Multi-dimensional (often seen with sliders and coded scatterplots)
    6. Tree (hierarchy)
    7. Network
  • Advanced filtering (partial search, boolean operations, etc) will be needed to help users search for data in ever-growing libraries
15google-maps-superjumbo
Google Maps shows an overview, permits zoom, and gives details when queried.

 

“Do Artifacts have Politics?” from The Whale and the Reactor: a Search for Limits in an Age of High Technology
Langdon Winner (1986)

  • Controversy: technical devices have political qualities, embody specific forms of power and authority.
  • “The theory of technological politics suggests that we pay attention to the characteristics of technical objects and the meaning of those characteristics.”
  • Two ways to show theory of technological politics:
    1. “Instances in which the invention, design, or arrangement of a specific technical device or system becomes a way of settling an issue in the affairs of a particular community.”

    2. “Cases of what can be called ‘inherently political technologies,’ man-made systems that appear to require or to be strongly compatible with particular kinds of political relationships”

  • The low bridges and overpasses of Long Island roads prevents buses from using the main roads–a purposeful device used from the 1920s through 1960s to keep inner city, lower class peoples out of Long Island.
  • The 1970s demonstrations of public works buildings that were not equipped to handle navigation by handicapped people; not a malicious use of technology, but that of mostly neglect.
  • A tomato picking machine, developed in 1940 California, could harvest a new style of hardier, less-tasty tomato faster and cheaper than hand-picking. A technology put approximately 32,000 out of work for a profitable 600 who could afford to keep their farms.
    • “What we see here instead is an ongoing social process in which scientific knowledge, technological invention, and corporate profit reinforce each other in deeply entrenched patterns, patterns that bear the unmistakable stamp of political and economic power.”

  • The technologies use to build our world situate some groups to be favored while some groups remain at various levels of awareness. It is often the most influential choices are made near the technologies inception, and those choices are ingrained in the technology making it difficult to divorce if issue arises.
  • “The automatic machinery of a big factory is much more despotic than the small capitalists who employ workers ever have been.” – Friedrich Engels, 1872

    • Authority in capitalism cannot be abolished, it can only be displaced and redistributed. Authority arises to cooperate teamwork and thus progress. A completely automatic factory is more authoritarian because it takes no input from workers, whereas small business leaders do.
  • Technology and advancement is a result a social structure existence, that without, would have never produced the technology in the first place.
  • “In many instances, to say that some technologies are inherently political is to say that certain widely accepted reasons of practical necessity–especially the need to maintain crucial technological systems as smoothly working entities–have tended.”
Harvest mechanization helps agriculture remain competitive
Despite the employment loss after it’s inception, this article says otherwise:

Although mechanization has reduced the number of labor hours for harvesting, overall employment for rice and processing tomatoes has risen due to increased production, and so have harvester operator wages.

 

Plot and ggplot2 with R-Studio

Data from Vision Problems in the U.S. provides estimates of the prevalence of eye-disorders in the US by state in adults 40 and older.

Here is the initial outputs of my work with R-Studio statistical software:

  1. When plot() is envoked on R-Studio, a matrix of plots are established by header columns. This particular function ran very slow, so I limited the data plotted to only entries from state: “OHIO”. Most noticeably, my values for vp (vision problem), age, race and sex are all categorical.
  2. I decided to further install the ggplot2 library for some more customizable plots. The color graph plots age against rate, with color determined by vp (vision problem). Each of my categorical variables has a “total/all” section that is not separated out from the individual state, race, gender or age data.
  3. I then attempted to use the box plot feature, which did not yield any additional insights.

Further steps to clearly visualize this data will address these problems:

  • What strategies are best to see averages for “all” categorical data alongside individual categorical data (e.g. female, or white, or 55-64 yrs)?
  • How can multiple categorical variables be presented at the same time (40-50yr Hispanic male vs. 51-60yr Black female vs. etc.)? Shapes, colors, other?
  • Should the data be presented as exploratory or tailored to meet a specific idea or perspective?
  • Which of these diseases/disorders result in a prescription for eyeglasses or vision correction?