Page 137 - Encyclopedia of Nursing Research
P. 137

104  n  DATA CoLLECTion METhoDS



           Loftus, 1993; Rozeboom, 1960; Walker, 1986),   quantitative analyses, computer programs for
           null hypothesis statistical testing still domi-  qualitative data analysis are merely aids for
   D       nates analytic approaches.               the tedious and error prone tasks of analysis.
              Some  of  the  articles  and  arguments   Using them still requires the investigator to
           about the limits of null hypothesis statistical   make the relevant and substantive decisions
           testing have led to more emphasis on the use   and  interpretations  about  codes,  categories,
           of confidence intervals. Confidence intervals   and themes.
           provide  more  information  about  our  find-  Although  quantitative  data  analysis
           ings, particularly about the precision of pop-  allows for statistical probabilistic statements
           ulation estimates from our sample data, but   to  support  the  investigator’s  interpretations
           they are based on the same null hypothesis   and  conclusions,  qualitative  data  analysis
           statistical testing logic that generates p val-  depends  more  exclusively  on  the  strength
           ues. Thus, confidence intervals are subject to   and  logic  of  the  investigator’s  arguments.
           the same issues with respect to Type 1 errors   nonetheless, both types of data analysis ulti-
           (rejecting the null when it is true) and Type 2    mately  rest  on  the  strength  of  the  original
           errors  (failing  to  reject  the  null  when  it  is   study design and the ability of the investiga-
           false).                                  tor  to  appropriately  and  accurately  execute
              increased  attention  and  sensitivity  to   the analytic method selected.
           factors that contribute to findings of statisti-
           cal significance has also led to more attention                Lauren S. Aaronson
           to power, sample sizes, and role of effect sizes
           (for substantive significance) for valid quan-
           titative data analysis. if the sample size is too
           small, the study may be underpowered and   Data ColleCtion MethoDs
           unable to detect an important finding even
           if it is there. Conversely, if the sample size is
           too large, the study may be overpowered and   in research, data are the pieces of informa-
           may result in statistically significant findings   tion that are gathered in an effort to address
           that  are  substantively  or  clinically  insignif-  a research question. Data collection typically
           icant. Either could be challenged on ethical   is  one  of  the  most  challenging  and  costly
           grounds, stressing the importance of appro-  steps  in  the  research  process.  Researchers
           priately  powering  studies  for  the  planned   make a number of decisions in designing a
           data analysis.                           data collection plan, and these decisions can
              in contrast to quantitative data analysis   have a profound effect on the quality of evi-
           which requires that the investigator assign a   dence that a study yields. nurse researchers
           numeric code to all data before beginning the   use a wide variety of methods for collecting
           analyses,  qualitative  data  analysis  consists   data, and these methods vary on a number of
           of coding words, objects, and/or events into   important dimensions.
           meaningful categories and/or themes as part   A  fundamental  dimension  involves
           of the actual data analyses. Because qualita-  whether  the  data  being  collected  are  quan-
           tive data analysis involves nonnumeric data,   titative or qualitative in nature. Quantitative
           there  are  no  statistical  probabilistic  tests  to   data yield information about a research var-
           apply to the coding of qualitative data.  iable in numeric form, ranging from simple
              Coding  of  qualitative  data  historically   binary values (e.g., 1 =  yes, 2 = no) to more
           has  been  done  manually,  but  computer   complex  numeric  expressions  (e.g.,  values
           programs  (e.g.,  QSR)  are  now  available  to   for the body mass index). To collect quanti-
           aid  the  investigator  in this  laborious  effort.   tative data, researchers use structured meth-
           however, as with the computer programs for   ods  and  formal  instruments  in  which  the
   132   133   134   135   136   137   138   139   140   141   142