Testing a Hypothesis
With Cross-Tabulation and Scatterplots
Using MicroCase CP

1. Run your tables

• For Cross-Tabulations: make the independent variable the column variable and the dependent variable the row variable

• For a Scatterplot: enter your variables properly as the independent and dependent variables

Note: for examples see the department's page on Table and Graph Format

2. Check to see if the differences in your table are statistically significant (the probability of chance findings being .05--five in one hundred--or less).

• for a cross tabulation, click on "statistics" and see if the "prob" value (prob=) is .05 or less For assessing the strength of a statistically-significant correlation, you may use the following rule-of-thumb:

  • if "V" is less than .1 (either positive or negative), consider it a weak relationship
  • if "V" is between .1.0 and 2.9 (either positive or negative), consider it a moderate relationship
  • if "V" is 3.0 or higher (either positive or negative), consider it a strong relationship

• for a scatterplot, see if the Pearson "r" below the scatterplot has one asterisk (.05 or less, but greater than .01) or two asterisks (.01 or less) or no asterisks (not significant). For assessing the strength of a statistically-significant correlation, you may use the following rule-of-thumb:

  • if "r" is less than .3 (either positive or negative), consider it a weak relationship
  • if "r" is between .3 and .6 (either positive or negative), consider it a moderate relationship
  • if "r" is higher than .6 (either positive or negative), consider it a strong relationship

3. Interpret the Findings

• If the probability of chance findings is greater than .05, the hypothesis must be rejected, regardless of any apparent pattern in the numbers in the table.  This is because there is a substantial chance that whatever differences are indicated by the table may be a result of random fluctuation in the sample.
• If the probability of chance findings is .05 or less, which means that the differences in your table are statistically significant and most likely not an outcome of chance, you must then check to see if the results are in the same direction as predicted by the hypothesis.
  • If they are, then the hypothesis is supported
  • If they are not, then the hypothesis must be rejected, even if the results are statistically significant.  (It may be possible to reformulate the hypothesis to bring it into line with the data, however.)

February 7, 2007