How to Choose Appropriate Statistic

Document Sample
How to Choose Appropriate Statistic Powered By Docstoc
					   If you see a picture of a
speaker on a slide, click on
 it to hear my commentary
         on that slide
 How to Use Bill Trochim’s Program for Choosing
            An Appropriate Statistic

• Probably the most complete system for choosing
  appropriate statistics when researcher has “raw” data can
  be found at:
• Bill Trochim – Cornell University – Center for Social
  Research – Statistics Selector
• http://trochim.human.cornell.edu/selstat/ssstart.htm
            Flaschner’s Procedure

• State Number of Variables
  – One
  – Two
  – Three
              One Variable
• Compute Descriptive Statistics
  – Depends on Level of Measurement


• Test if Statistic is significantly different
  from corresponding population Parameter
  – One Sample
  – Depends on Level of Measurement
Verification in Choosing
          Tests




                           Verification for
             Two Variables
• Calculate Measure of Association
  – Probable Change in Y occurs due to change in
    X
Measures of Association



                      Measures of Association
           Test for Differences
• Between
  –   Frequencies      – Variances
  –   Proportions      – Standard Deviations
  –   Medians          – Correlations
  –   Means            – b’s
                       B’s
    Difference Between Means
• Size of smallest sample
• Test of independence
  – Contingency Coefficient - C
• Test for differences between variances
Verification in Choosing
          Tests




                           Verification in
        Possible Statistics Using Excel’s Tools/Data
                     Analysis (Add In)
Anova: Single Factor
Anova: Two-Factor With Replication
Anova: Two-Factor Without Replication
Correlation                      Regression
Covariance                       Sampling
Descriptive Statistics           T-Test: Paired Sample for Means
Exponential Smoothing            T-Test: Two-Sample Assuming
F-Test Two-Sample for Variances         Equal Variances
Fourier Analysis                 T-Test: Two-Sample Assuming
Histogram                               Unequal Variances
Moving Average                   Z-Test: Two Sample for Means
Random Number Generation
Rank and Percentile
Possible Statistics Using SPSS
          Verification
• What path do I follow to be
  allowed to calculate the
  statistics in which I have an
  interest?