2. Measurements Choices Determine Analysis Options Choice of measure occurs in research design Measures yield data at a particular scale (level) of measurement (Nominal, Ordinal, Interval, Ratio) Interval or Ratio data can be “recoded” to create groups (e.g., age groups, income ranges) Grouped data cannot be expanded to create Interval or Ratio data Scale (Level) of Measurement determines which techniques are appropriate.
4. Matching Variability Measure to Data Range and Interquartile Range (IQR) use only a few scores. Standard deviation and variance use the value of each score in the data set Range and IQR are related to Median Standard Deviation and Variance are related to the Mean.
5. Purpose of a Graph A visual presentation of data Relationships & comparisons are visual Less daunting to some than tables of numbers Allows some artistry and creativity Accuracy is important Characteristics of data Measurement choices in design determine analysis choices later Scale (level) of measurement determines which graphs can be used Nature of the particular data set is also important
6. Graphs for Complex Data The Future of Food. (2008) WiredMagazine 16:11 From ChoiceRanker website via JunkCharts blog at http://junkcharts.typepad.com/junk_charts/2008/07/its-raining-colors-here-too.html
16. BAR CHART: the Good Area of bars combined is 100% Area of each bar is proportional to its percent of total Bars do not touchbecause categoriesare discrete. Many variations; this is the most simple.
19. PICTOGRAPH: the Ugly Elements of unequal size Just heads of some kids All children are playing except those from China – subtle racism
20. BAR CHART – problems to consider:area, color – & why is that jogger there?
21. PIE CHART: the Good Area of pie = 100% Wedge is proportional to percentage of cases Labels show count or percent Ten slices is the maximumto remain clear
22. PIE CHART: the Badcharts confuse or obscure the pattern in the data
23. Graphs for Continuous Data (sometimes Ordinal) Graph shows continuity of the construct Histogram: bars that touch at real limits Line graph: covers range (a.k.a. Frequency Polygon) Horizontal axis goes from low to high Intervals shown for Interval or Ratio data Some ordinal data also graphed this way(e.g., strongly agree, agree, slightly agree, etc)
24. HISTOGRAM: the Good Bar width is a rangeof scores or the reallimits of scores. Ranges equal width Labels show mid-point or real limits Low scores on left, high scores on right
25. HISTOGRAM: the Bad Ranges of data Unequal Indeterminate Spacing of “bars” is unequal. Water, sky, umbrellaall detract from graph
26. Line Graphs / Frequency Polygon Same requirements as histogram. If more than one line,legend or labels are needed. More than four or fivelines can be hard tointerpret from SRB Documentary. (2008). Demographic Winter: the Decline of the Human Family at http://www.demographicwinter.com/index.html
27. LINE GRAPH: the Bad Why is the headline “Steady growth” for this graph? Hint: check the axis values If it is growth, is it steady ? Hint: how did each of the three variables change from 1988 to 1989.
The measures we looked at so far draw only two data points, rather than the value of each score or value in the data set. They do not relate each score directly to the measure of central tendency. The next methods we consider, the standard deviation and the variance, are able to accomplish both of those goals. They are presented in the next lecture. This table, which will appear again at the end of lecture 2, can be used in deciding which measure of variability to use (if you are the researcher) or to evaluate whether the measure used in a research report was, in fact, the best choice for that situation.