Scatterplotting


Return to Previous Page

A scatterplot is a graphing of each individual outcome (y-axis, dependent or continuous variable) plotted against the associated independent variable (x-axis, continuous or categorical variable). This shows if a relationship of an independent variable with an outcome variable exists.

Here are the steps in producing the scatterplots for Bangladesh:

  1. Open bdeshc.sav.
  2. Click in the menu on Graphs, Scatter, Simple, and Define
  3. In the left hand variable list, scroll down and select the dependent variable acprvfm (for Bangladesh) and click the arrow button to place them into the Y-axis box.
  4. Now select the independent variable of interest for this scatterplot  wathoth (% using other water sources),wathcat (water source categories for high and low, see if it depicts a similar relationship as wathoth), latcat (sanitation category) and place it in the X-axis box with a click on the arrow button, and click OK.
  5. Double click on the graph that results, to make it editable (a new smaller window will appear with an editable graph)
  6. Click on Chart, Options and place a check in the Fit Line, Total box. This will place a linear regression line on the chart to show the association between the variables.
  7. Check if the slope goes in the expected direction.

 

BANGLADESH (district level):

wpe5.jpg (16322 bytes)

INTERPRETATION: As a continuous variable, the prevalence of households using other (poor) water sources are scatterplotted here against malnutrition, which does show the relationship one would expect.  There is higher malnutrition (low arm circumference) prevalence in those areas where use of poor sources of water high and vice versa.  This variable for poor source of water has also been categorized into high and low prevalence of poor water source (variable called wathcat) and is scatterplotted against malnutrition below.  A similar relationship is seen.

wpe3.jpg (15335 bytes)

INTERPRETATION: Using the categorical variable for water source, you also see a positive slope that indicates that those with access to safe water source (0) have a lower prevalence of low arm circumference than those without access to safe water (1).  This gives a very similar result as that with the continuous variable for prevalence of access to water (district level data) in the previous scatterplot.  We will use the categorized variable in the regression analysis exercise.  Now, take a look at the scatterplot for Low AC and latrine safety.

wpe4.jpg (16404 bytes)

INTERPRETATION: The variable for access to safe latrine at the district level has been categorized as low (0) and high (1) or bad and good respectively.   There is also a clear relationship in the scatterplot between low access to latrine and high prevalence of malnutrition (low arm circumference) in comparison to those areas with better access to safe latrines and a lower prevalence of malnutrition.  This is in the expected direction.

 

KENYA (individual level):

  1. Open keast4j.sav.
  2. Click in the menu on Graphs, Scatter, Simple, and Define
  3. In the left hand variable list, scroll down and select the dependent variable waz and click the arrow button to place them into the Y-axis box.
  4. Now select the independent variable of interest for this scatterplot dbadro (roofing in categories of good and bad- Kenya),  or dlowedn (education attainment in categories of good and bad- Kenya) and place it in the X-axis box with a click on the arrow button, and click OK.
  5. Double click on the graph that results, to make it editable (a new smaller window will appear with an editable graph)
  6. Click on Chart, Options and place a check in the Fit Line, Total box. This will place a linear regression line on the chart to show the association between the variables.
  7. Check if the slope goes in the expected direction.

wpe2.jpg (16589 bytes)

INTERPRETATION: The variable for roofing quality is categorized as bad (grass/ thatch) and good (corrugated iron), and the association with the outcome variable of WAZ score does not appear to have a significant slope.  There is a slightly lower waz score associated with those that have grass/ thatch roof than those with corrugated iron, which is in the expected direction.

 

wpe1.jpg (17594 bytes)

INTERPRETATION: The variable for education level has been categorized as a dummy variable (dlowedn) and shows an association with waz score when scatterplotted.  This supports the expected association, those with low education have lower nutrition status on average; therefore, it is likely a good candidate for running in a regression analysis.

Return to Top of Page