Transcript Unit 4 Data Analysis

Data Analysis in Quantitative
and Qualitative Research
1
Quantitative
Data
Analysis
2
• Data collected in research studies needs to be
systematically analyzed to determine trends and patterns
of relationships.
• Statistical procedures are used to do this in quantitative
research
3
Levels of Measurement
• A system of classifying measurements according to the
nature of the measurement and the type of mathematical
operations (statistics) which can be used to measure it
4
Levels of Measurement
•
•
•
•
Nominal measurement
Ordinal measurement
Interval measurement
Ratio measurement
5
Levels of Measurement
• Nominal measurement
– Uses numbers simply to categorize characteristics
– Method of sorting data
– Provides information only about categorical
equivalence and non-equivalence
– Can not be treated mathematically (can’t be quantified)
– Lowest level of measurement
• i.e. gender, blood type, nursing speciality
• May assign males to be 1, females to be 2
6
Levels of Measurement
• Ordinal measurement
– Ranks objects based on their relative standing on a
specific attribute
– Rank orders
– Limited ability for mathematical formula (quantifiable)
•
•
•
•
i.e. lightest person to heaviest person
Tallest person to shortest
Totally independent to total dependent
Level of education – high school, BN, MN
7
Levels of Measurement
• Interval measurement
– Ranking of objects on an attribute but also able to
specify the distance between those objects
– Used for statistical procedures
• i.e. scholastic testing
• Psychological testing
8
Levels of Measurement
• Ratio measurement
– Ratio scales have a rational, meaningful zero
– Provide information about the absolute magnitude of
the attribute
– Used for statistical procedures
– Highest level of measurement
• Blood pressure
• Weight (200 pds is twice as heavy as 100 pds)
• Intake / Output
9
What is Statistics
• Statistical procedures enable the researcher to
summarize, organize, interpret, and communicate
numeric information
10
Statistics
• Classification
– Descriptive statistics
– Inferential statistics
11
Descriptive Statistics
• Are used to describe and synthesize data
– i.e. averages and percentages
– Parameters
• A characteristic of a population
• Data calculated from a population
– i.e. mean age of all Canadian citizens
– Statistic
• Data calculated from a sample
• An estimate of a parameter
12
Descriptive Statistics
• Most scientific questions are about parameters,
researchers calculate statistics to estimate them
13
Descriptive Statistics
• Methods researchers use to make sense out of
descriptive data
– Frequency distributions
– Central tendency
– Variability
14
Descriptive Statistics
• Frequency Distributions
– Imposes order on numeric data
– Systematic arrangement of numeric values from lowest to
highest
– Displays percentage of the number of times each value was
obtained
15
Descriptive Statistics
• Frequency Distributions
– Can be displayed in
• Frequency polygon
– Graphically - scores on horizontal line, frequency (percentages) on vertical
line
100%
80%
60%
40%
20%
0%
North
West
East
1st Qtr
2nd Qtr
3rd Qtr
4th Qtr
16
Descriptive Statistics
• Central Tendency
– Seeks a single number that best represents the whole
distribution
– Describes only one variable
• Mode
• Median
• Mean
17
Descriptive Statistics
– Mode
• The number that occurs most frequently in a distribution
• The most popular value
• Used for describing typical or high-frequency values for
nominal measures
• i.e. 5 10 6 10 9 8 10 (10)
18
Descriptive Statistics
– Median (Md)
• The point in the distribution where 50% of cases are above
and 50% of cases are below
• The midpoint in the data
• Insensitive to extreme values
• Used with highly skewed distributions
• i.e. 2 2 3 3 4 5 6 7 8 9
(4.5)
19
Descriptive Statistics
– Mean
• Is equal to the sum of all values divided by the number of
participants
• The average
• Affected by the value of every score
• Used in interval or ratio-level measurements
• The mean is more stable than the median or mode
• i.e. 5 4 3 2 1 15/ 5 = (3)
20
Descriptive Statistics
• Variability
– Variability is the spread of the data from the mean
– The variability of two distributions could be different
even when the means are identical
• i.e. could have a homogenous group with scores all clustered
together with the same mean as a heterogeneous group
where scores are variable
– Need to know to what extent the scores in a
distribution differ from one another
21
Descriptive Statistics
• Variability
– Used to measure the variability, the differences in
dispersion of data
• Range
• Standard Deviation
22
Descriptive Statistics
• Range
– Is the highest score minus the lowest score in a
distribution
– Not a very stable method
– Only based on two scores
» i.e. highest score 750, lowest score 250 range
(500)
23
Descriptive Statistics
– Standard Deviation (SD)
• Calculated on every value in a distribution
• Summarizes the average amount of deviation of values from
the mean
• Most widely used measurement to determine variability of
scores in a distribution
24
Descriptive Statistics
– Standard Deviation (SD)
• The SD represents the average of deviations from the mean
• Indicates the degree of error when a mean is used to
describe data
• In normal distributions there are three standard deviations
above and below the mean
– i.e. 68% of cases fall within 1 SD above and below the mean
– 95% of cases fall within 2 SD above and below the mean
– 4% of cases fall more than 2 SD from the mean
25
Standard Deviation Graph
26
Descriptive Statistics
• Bivariate Descriptive Statistics
– Bivariate (two variable)
– Describes relationships between two variables
– Uses
• Contingency Tables
• Correlation
27
Descriptive Statistics
– Contingency Tables
• The frequency of two variables are cross-tabulated
• Used for nominal or ordinal data
– i.e. look at both gender and non-smokers (two variables)
28
Descriptive Statistics
– Correlation
• Most common method of describing the relationship between
two variables
• Calculate the correlation coefficient
– Describes the intensity and direction of the relationship
– A formula that determines how perfect the relationship is
– Positive relationship (+1.00), negative relationship (-1.00), no
relationship (0)
– When two variables are positively correlated this means high values
on one variable are associated with high values on the other variable
– i.e. height and weight
– Tall people tend to weight more than short people
29
Inferential Statistics
• Entails formulas that provide a means for drawing
conclusions about a population from the sample
data
30
Inferential Statistics
• Sampling Distributions
– When estimating population characteristics, you need
to obtain representative samples
– Probability sampling is best
– Inferential statistics should use only probability
sampling
31
Inferential Statistics
• Sampling error
– As it is not possible to obtain a sample that is identical to the
population, a slight error is assumed
– The challenge for researchers is to determine whether sample
values are good estimates of population parameters
• Standard error of the mean
– The sample means of the distribution contain some error in their
estimates of the population
– The smaller the standard error the more accurate are the means
as estimates of the population value
32
Inferential Statistics
• The more homogenous the population the more
likely it is that the results from a sample will be
accurate
• The larger the sample size, the greater is the
likelihood of accuracy as extreme cases will cancel
each other out
33
Inferential Statistics
• Uses two major techniques
– Estimation of Parameters
– Hypothesis testing
34
Estimation of Parameters
• Estimation procedures (estimation of parameters)
are used to estimate a single population
characteristic
• Not used often as researchers are more interested
in relationships between variables
– i.e. oral temperature
35
Hypothesis Testing
• Hypothesis
– A prediction about relationships between variables
• Null Hypothesis
– States that there is no relationship between the independent and
the dependent variables
– Determining that the null hypothesis has a high probability of
being incorrect, lends support to the scientific hypothesis
– Rejection of null hypothesis is accomplished through statistical
tests
– Rarely stated in research reports
36
Hypothesis Testing
• Provides objective criteria for deciding whether
hypotheses should be accepted as true or rejected
as false
• Assists researchers decide which results are likely
to reflect chance differences and which are likely
to reflect true hypothesized effects
• Researchers assume that the null hypothesis is
true and then gather evidence to disprove it
37
Hypothesis Testing
• Type l and Type ll Errors
– Researchers decide whether to accept or reject the null
hypothesis by determining how probable it is that
observed group differences are due to chance
38
Hypothesis Testing
• Type l Error
– Rejecting the null hypothesis when it is true
• i.e. concluded that the experimental treatment was effective
when in fact the group differences were due to sampling
error or chance
• Type ll Error
– Accepting a false null hypothesis
• i.e. concluded that the observed differences were due to
random sampling fluctuations when in fact the experimental
treatment did have an effect
39
Hypothesis Testing
• Level of Significance (alpha) is:
– The risk of making a Type l error, established by the researcher before
statistical analysis
– The probability of rejecting the null hypothesis when it is true
– .05 level
» accept risk that out of 100 samples, a true null hypothesis would
be rejected 5 times (5%)
» Accept that in 95 of 100 cases (95%), a true null hypothesis
would be correctly accepted
– .01 level
» Accept that 1 out of 100 samples, a true null hypothesis would
be rejected (1%)
» Accept that in 99 of 100 cases, a true null hypothesis would be
rejected (99%)
– .001 level
» Accept that 1 out of 1000 samples, a true null hypothesis would40
be rejected
Hypothesis Testing
• Level of Significance (alpha)
– The minimal acceptable alpha level for scientific
research is .05
– Lowering the risk for Type l error increases the risk of a
Type ll error
– Can reduce the risk of Type ll error simply by
increasing the sample size
41
P Values (the probability value)
• Level of significance is sometimes reported as the actual computed
probability that the null hypothesis is correct (based on research
results)
– Can be reported as falling below or above the researcher’s significance
criterion < or >
– P = .09 (9 out of 100 chance that observed differences between groups could
be by chance)
– P < .05 (5 out of 100 chance that observed differences could be by chance)
– P < .01 (1 out of 100 chance that observed differences could be by chance)
– P < .001 (1 out of 1000 chance that observed differences could be by
chance)
42
Hypothesis Testing
• Researchers reporting the results of hypothesis
tests state that their findings are statistically
significant
– Significance means that the results were most likely
not due to chance
– The statistical findings supported the hypothesis
– Nonsignificant results means that the results may have
been the result of chance
43
What Does This All Mean?
• If a researcher reports that the results have statistical
significance this means that based on statistical tests, the
findings are probably valid and replicable with a new
sample of participants
• The level of significance is how probable it is that the
findings are reliable and were not due to chance
• If findings were significant at the .05 level this means that:
– 5% of the time the obtained results could be incorrect or due to
chance
– 95% of the time similar results would be obtained if you did
multiple tests, therefore the results were probably not due to 44
chance
Tips on Reading Statistical
Information
• Information reported in the results section:
– Statistical information
• Enables readers to evaluate the extent of any biases
– Descriptive statistics
• Overview of participant's characteristics
– Inferential statistics
• What test was used
• Further statistical information
45
Interpreting Study Results
• Must consider
–
–
–
–
–
The credibility and accuracy of the results
The meaning of the results
The importance of the results
The extent to which the results can be generalized
The implication for practice, theory, or research
46
Interpreting Study Results
– The credibility and accuracy of the results
•
•
•
•
•
Accurate and believable
Based on analysis of evidence not personal opinions
External evidence comes primarily from the body of prior research
Research methods used
Note limitations
– The meaning of the results
• Analyzing the statistical values and probability levels
• It is always possible that relationships in the hypothesis were due to
chance
• Non-significant findings represent a lack of evidence for either truth or
falsity of the hypothesis
47
Interpreting Study Results
– The importance of the results
• Should be of importance to nursing and the healthcare of the
population
– The generalizability of the results
• The aim is to gain insights for improvement in nursing
practice across all groups and within all settings
– The implications of the results
• How do the results affect future research
• How do the results affect nursing practice
48
• Analyzing Qualitative Data
49
Qualitative Analysis
• Includes data from:
• Loosely structured
narrative materials
from interviews
• Field notes or
personal diaries from
personal observation
50
Qualitative Analysis
• Purpose of Analysis
– To organize, provide
structure, and elicit
meaning from research
data
51
Qualitative Analysis
• Difficult to analysis data as:
– Qualitative research does not have systematic rules for
analyzing data
– A lot of work, very time consuming
– Difficult to reduce data to report findings
52
Qualitative Analysis: Styles
• Analysis Styles:
–
–
–
–
Quasi-statistical analysis style
Template analysis style
Editing analysis style
Immersion/crystallization analysis style
53
Qualitative Analysis: Styles
– Quasi-statistical analysis style
• Views narrative data for particular words or themes
• Collects information that can be analyzed statistically
– Template analysis style
• Uses a template to determine which narrative data are
analysed
– Editing analysis style
• Codes and sorts data after the researcher has identified
meaningful segments
– Immersion/crystallization analysis style
• Total immersion in and reflection on the materials
54
Qualitative Analysis: Process
• Analysis Process:
– Comprehending
• Completed when saturation has been attained
– Synthesizing
• Putting the data together and making sense out of it
– Theorizing
• Sorting the data and developing possible theories
– Recontextualizing
• Developing theory further so it could be applied to other
settings or groups
55
Qualitative Analysis: Procedures
• Organizing the Data:
– Classify and index the material, identifying the concepts
– Reductionist – data is converted to smaller, more manageable
units that can be retrieved and reviewed
– Develop categorization scheme, code data according to
categories
– Constant comparison –collected data is compared continuously
with data obtained earlier
56
Qualitative Analysis: Procedures
• Coding the Data:
– Data are coded into appropriate categories
– Manual Methods
• Develops files to sort concepts
• Cuts out concepts from narratives for filing in concept folder
– Computer Programs
• Indexes and retrieves data
• Also can analyze data
57
Qualitative Analysis: Procedures
• Analyzing the Data
– Constructionist – putting pieces together into a meaningful
pattern
– Begins with the search for relationships and themes in the data
– Validation that the themes are accurate representation of the
phenomenon
• Investigator triangulation – using more than one person to
interpret/analyze the data
• Member checks – share preliminary findings with informants to get their
opinions
– Unites themes to develop theory
58
Qualitative Analysis
• Grounded Theory Analysis
– Generating theories and conceptual models from data
– Constant comparison – simultaneously collects, codes
and analyzes data
– Uses coding – open and selective
59
Qualitative Analysis
• Phenomenological Analysis
– Describes the meaning of an experience, identifies
major themes
• Qualitative Content Analysis
– The analysis of the content of narrative data to identify
prominent themes and patterns among the themes
60
Interpreting Qualitative Results
•
•
•
•
•
Credibility
Meaning
Importance
Transferability
Implications
61
Reference
• Loiselle, C. G., Profetto-McGrath, J., Polit, D. F., &
Beck, C. T. (2011). Canadian essentials of nursing
research. (Third Edition). Philadelphia: Lippincott,
Williams & Wilkins.
62