Nursing Data Analysis In Research

Data Analysis

    Data analysis
is a systematic method of examining data gathered for any research
investigation to support conclusions or interpretations about the data.
Although applicable to both qualitative and quantitative research data analysis
is more often associated with quantitative research. 

    Quantitative data analysis
involves the application of logic and reasoning through the use of statistics,
an applied branch of mathematics, to numeric data. Qualitative data analysis
involves the application of logic and reasoning, a branch of philosophy, to
nonnumeric data. 

    Both require careful execution and are intended to give
meaning to data by organizing disparate pieces of information into
understandable and useful aggregates, statements, or hypotheses.

Specific Statistical Tests To Be Used

    Statistical
data analysis is based in probability theory and involves using a number of
specific statistical tests, or measures of association between two or more
variables. 

    Each of these tests or statistics (e.g., t, F, B, z, o, 7, etc.) has
a known distribution that allows the calculation of probability levels for
different values of the statistic under different assumptions that is, the test
(or null) hypothesis and the sample size, or degrees of freedom. Specific tests
are selected because they provide the most meaningful representation of the
data in response to the research questions or hypotheses posed. 

    The selection
of specific tests, however, is restricted to those for which the available data
meet certain required assumptions of the tests. For example, some tests are
appropriate for (and assume) nominal data, others assume ordinal data, and
still others assume an interval level of measurement. 

    Although each test has
its own set of mathematical assumptions about the data, all statistical tests
assume random sampling.

Computer Programs In Data Analysis

    Several
statistical computer programs (e.g., SPSS, SAS, LISREL, EQS) can aid the
investigator with the tedious and complex mathematical operations necessary to
calculate these test statistics and their sampling distributions. 

    These
programs, however, serve only to expedite calculations and ensure accuracy.
Because the investigator must understand the computer programs to use them
appropriately, there is a hidden danger in the ease with which one may execute
such programs. 

    For valid data analysis, the investigator must fully
understand the underlying statistical procedures and the implied assumptions of
these tests in order to apply them appropriately.

Hypothesis Statistical Data Analysis

    The logic of
null hypothesis statistical data analysis is one of modustollens, denying the
antecedent by denying the consequent. That is, if the null hypothesis is
correct, our findings cannot occur but our findings did occur, so the null
hypothesis must be false. 

    However, J. Cohen (1994) and others have
convincingly argued that, by making this reasoning probabilistic for null
hypothesis statistical testing, the original syllogism is invalidated. 

    Moreover, for decades scientists from different disciplines have questioned the
usefulness and triviality of null hypothesis statistical testing (see Labovitz,
1970; LeFort, 1993; Loftus, 1993; Rozeboom, 1960; Walker, A. M., 1986, for
examples from sociology, psychology, public health, and nursing). 

    Consequently,
increased attention to the factors that contribute to findings of statistical
significance is warranted and power, effect sizes (for substantive
significance), sample sizes, and confidence intervals are receiving in- creased
attention in quantitative data analysis.

Difference In Qualitative And Quantitative Data Analysis

    In contrast to
quantitative data analysis, which requires that the investigator assign a
numeric code to all data prior to beginning the analyses, qualitative data
analysis consists of coding words, objects, or events into coherent or
meaningful categories or themes as part of the actual data analyses. 

    Also,
because qualitative data analysis involves nonnumeric data, there are no
statistical probabilistic tests to apply to their coding.

Historical Perspective of Data Analysis

    Historically,
qualitative data coding has been done manually, but more recently computer
programs (e.g., NUDIST) have been developed to aid the investigator in this
laborious effort. 

    However, as with the computer programs for quantitative
analyses, those for qualitative data analysis are merely aids for the tedious
and error prone tasks of analysis. Using them still requires that the investigator
make the relevant and substantive decisions and interpretations about codes,
categories, and themes.

    Quantitative
data analysis allows for statistical probabilistic statements to support the
investigator’s interpretations and conclusions. Qualitative data analysis
depends more exclusively on the strength and logic of the investigator’s
arguments. 

    Nonetheless, both types of data analysis ultimately rest on the
strength of the original study design and the ability of the investigator to
appropriately and accurately execute the analytic method selected.