Unit 7

Wildemuth Chapters 29-37

Methods for Data Analysis

Content Analysis
Interested only in analyzing the content in relation to the hypothesis

can be used to predict the outcome of a message

Manifest vs latent content
manifest - obvious like a specific word or a color
latent - conceptual not directly observed

Units for analysis

• sampling units
• recording units
• physical
• conceptual
• temporal

Coding scheme development

pg. 300 - Coding - reducing the entire content of the message into quantitatively analyzable data that only describes the variables which you are studying

coding should be done by more than one person to avoid bias, but this is not often done and/or successful

Analysis of coded data

frequencies of each category

Qualitative Analysis of Content

examines meanings, themes, and patterns

3 approaches

• conventional qualitative content analysis
• direct content analysis
• summative content analysis

Process of qualitative content analysis pg. 312-313

• Prepare the data
• should the whole thing be transcribed
• should verbalizations be transcribed literally or in summary
• should observations be transcribed
• Define the unit of analysis
• Develop the categories and coding scheme
• Test your coding scheme on a sample
• Code all the text
• Asses coding consistency
• Draw conclusions from the coded data
• Report your methods and findings

To improve creditability pg. 313

• prolonged time in field
• persistent observation
• triangulation
• negative case analysis
• checking interpretations against raw data
• peer debriefing 
• member checking

transferability - how well the working hypothesis can be applied to another context

Discourse Analysis

Most studies see social phenomena as the products of discourse. pg. 321

Research focusing on how and why the discourse is used

role of speech is to construct the speakers' social worked pg. 321

Analytic Induction

deductive reasoning - given one broader set of facts is true, another more specific fact must also be true

inductive reasoning - examining specific facts and applying those truths to a more general conclusion.

analytic induction

• formalized method
• used for refining theory or hypothesis from data
• used for defining variables or phenomena

used for

• ethnographic observation
• participant observation
• semi-structured and unstructured interviews

Process of analytic induction pg. 330-331

• formulate rough definition of phenomenon you are to explain
• develop hypothetical explanation of the phenomenon
• choose cases and study them - include likely negative cases
• negative case requires 1 of 2 options
• redefine the phenomenon
• reformulate the hypothesis

one challenge is knowing when to stop narrowing phenomenon

Descriptive Statistics

measures of central tendency - how values of a variable cluster together

measure of dispersion - how values for a variable disperse from each other

4 levels of measurement

• nominal variables - categories, no true numerical value can be given for them
• ordinal variables - values can be ranked in order
• interval variables - uniform distance between possible values, can also be ordered
• ratio-level variables - ordered, uniform distance, and have a true zero.

measures of central tendency

• mean - often used, but can be skewed by extremes
• median
• mode - only option for nominal data, actual score from observation

will use one number to describe your entire data set

range - dance from the lowest to the highest number in the set

inter-quartile range - the difference between the value at the 25% and 75% so representative of the range of 50% of the data

when using mean report the standard deviation

confidence intervals - level of confidence you have in the range you set for the mean of the variables

Frequencies, Cross-tabulation, and the Chi-square Statistic

frequency distribution - a table displaying the count of cases in a particular category within a particular variable

calculate relative frequency only when sample size is large

two-way table shows the overlap between two variables, number of instances that satisfy both

3 guidelines for tables pg. 349

1. consider order for rows and columns
2. round numbers off to a point they will be easily interpreted
3. provide a meaningful summary for each row and column

The chi-square statistic

"measures the difference between what is observed and what would be expected in the general population" pg. 349

warnings about chi-square

1. does not work well with low numbers
2. when using a two-by-two column table use Fisher's exact test
3. with too large of a sample size may see relationships that aren't meaningful
4. chi-square only tells you a relationship is likely to exist, not that it does exist

Visualization of data set

graphs and charts are helpful to visualize the data, but need to be accompanied by the actual statistics

• pie charts
• bar charts
• histograms
• line graphs
• box plots

Analyzing Sequence of Events
in LIS most often applied to studying information seeking behaviors and analyzing searching and browsing behaviors.

Working with sequential data pg. 361

• capture a record of the sequence of the behavior being studied
• transform raw data into a form that can be analyzed
• analyze the sequences so that they can be looked at in comparison to the original research question

Coding the data

• select the events that are of interest
• chunk like events together
• assign codes to the events

Analysis

2 categories

• focused on the transitions from one event to another
• based on optional matching

Correlation

Examine the relationship between two variables

strength of relationship is based on the absolute of the correlation value can range from 0 to 1.

correlation does not necessarily mean causation.

Comparing Means:  t Tests and the Analysis of Variance

actual effects from chance effects

two steps to dealing with random variation pg. 383

1. define an appropriate sample
2. use and interpret the appropriate statistics

The t test

3 basic ideas pg. 384

1. large differences are more meaningful than small ones - the bigger the difference between two groups the larger the value of t
2. smaller variances are more reliable than larger ones - smaller variance of responses within groups means a larger value for t.
3. larger samples are more reliable than smaller ones - the larger the size of the sample the larger the value of t.

t  = the difference between the means divided by the standard error

t allows you to determine p (the probability that the difference between the two means is caused by chance).  t and p  are inversely proportional.

Comparing multiple groups: ANOVA
used for comparing more than two groups

still uses means like the t test.

Connaway & Powell Chapter 9: Analysis of Data

statistics allow you to determine the reliability of the conclusion you draw from your data

theoretical and applied statistics

role of statistics
4 basic purposes pg. 261-262

• indicate the central point around which the mass of data revolves
• show how broad and diverse is the spread of data
• how closely related certain features within the mass are
• may indicate if there is a likelihood of a causal relationship between the facts

Steps involved in statistical analysis

• establishment of categories
• coding the data
• analysis

descriptive statistics - pie graph, bar graph

inferential statistics - scatter graph