Data Analysis

Data Analysis

The graphs and charts are used to classify data and keep everything organized in our lives. It is used everywhere, such as- selling and purchasing shares, company financial records and even our class list makes use of tables. Surveys, mathematical tables and many others make use of data analysis through graphs and tables. Making a shopping list and using graphs to keep track of expenses are more uses of that form of data analysis.

We do not agree with that person’s point of view as different types of graphs have different uses and variant. Some examples are listed below.

Area Chart
Figure 1: Area chart
Use it to...
  • Display over time (or any other dimension):
    • How a set of data adds up to a whole (cumulated totals)
    • Which part of the whole each element represents
  • Percentage: The sum always represents 100% (relative scale)
  • Cumulative: The sum can vary according to the elements (absolute scale)

Column/Bar Chart
Use it to...
  • Present few data over a nominal (e.g. countries, testing conditions, ...) or interval scale (e.g. time); useful for comparisons of data
Do not Use it for...
  • Comparisons: Better use one-dimensional scatterplots, because these are not dominated by bars or columns.
  • Larger data sets: Use line charts.
Selecting Bars or Columns
  • Use analogy as a selection criterion, if applicable; when in doubt, use columns
  • Use a horizontal bar chart if the labels are too long to fit under the columns
  • Multiple Column/Bar Chart: Use it to present data rows for several variables
  • Side-by-Side Chart: Use it to (1) show contrasting trends between levels of an independent variable, (2) if comparisons between individual pairs of values are most important; do not use for more than two independent variables


Figure 2: Multiple column chart (left), side-by-side chart (right)

Segmented Column/Bar Chart
Other Names: Divided or stacked column/bar chart
Figure 3: Segmented column chart (relative values)
Use it to...
  • Present a part-whole relation over time (with accurate impression, see below)
  • Show proportional relationships over time
  • Display wholes which are levels on a nominal scale
Segmented column/bar charts are more accurate than pie chart, because distances can be more accurately estimated than areas.

Frequency Polygon, Histograms
Figure 4: Histogram as frequency distribution
  • Polygon: Connects data points through straight lines or higher order graphs
  • Histogram: Columns/bars touch; useful for larger sets of data points, typically used for frequency distributions
  • Staircase Chart: Displays only the silhouette of the histogram; useful for even larger sets of data points, typically used for frequency distributions
  • Step chart: Use it to illustrate trends among more than two members of nominal or ordinal scales; do not use it for two or more variables or levels of a single variable (hard to read)
  • Pyramid histogram: Two mirror histograms; use it for comparisons

Line Chart
Figure 5: Line chart
Use it...
  • To display long data rows
  • To interpolate between data points
  • To extrapolate beyond known data values (forecast)
  • To compare different graphs
  • To find and compare trends (changes over time)
  • To recognize correlations and covariations between variables
  • If the X axis requires an interval scale
  • To display interactions over two levels on the X axis
  • When convention defines meaningful patterns (e.g. a zigzag line)
Line graphs may consist of line or curved segments:
  • Lines: Use straight lines to connect "real" data points
  • Curves: Use curves to represent functional relations between data points or to interpolate data
Do not Use it...
  • If the X axis has non-numeric values
  • Graph with double-logarithmic or half-logarithmic scale divisions
  • Graph with variance bars, stock charts (High/Low/Close) etc.

Pie Chart
Figure 6: Pie chart
Use it to...
  • convey approximate proportional relationships (relative amounts) at a point in time
  • compare part of a whole at a given point in time
  • Exploded: emphasize a small proportion of parts
Do not Use it ...
  • For exact comparisons of values, because estimating angles is difficult for people.
  • For rank data: Use column/bar charts in this case; use multiple column/bar charts for grouped data
  • If proportions vary greatly; do not use multiple pies to compare corresponding parts.
  • Pie charts cannot represent values beyond 100%.
  • Each pie chart is valid for one point in time only.
  • Pie charts are only suited to presenting quite a few percentage values.
  • Angles are harder to estimate for people than distances; perspective pie charts are even harder to interpret.



Figure 7: One-dimensional scatterplot (left), two-dimensional scatterplot (right)
  1. One-dimensional scatterplot: Data point are drawn above a baseline (as in column/bar charts). Here the data points are not connected but remain isolated data points.
  2. Two-dimensional scatter plot: Shows correlation between two data sets. This chart type has two dependent variables: One is plotted along the X axis, the other along the Y axis; the independent variable is the intersection of both dependent variables, realized as a data point in the diagram.
Use it to...
  • Show measurements over time (one-dimensional scatterplot)
  • Convey an overall impression of the relation between two variables (Two-dimensional scatterplot)
Do not Use it for...
  • Determining and comparing trends, interpolation, extrapolation, recognition and comparison of change rates
  • More than one independent variable: Avoid illustrating more than one independent variable in a scatter plot

Loh ChengNgee, Khit Sue Lun, Lai Ziying

No comments:

Post a Comment