Means between 3 or more groups

Means between 3 or more groups

by Isaac -

Hey, I think this should be simple, but it's apparently not (for me)....

I'm trying to compare means (of age) between 3 different groups (a categorical variable with 3 outcomes).

I tried an ANOVA, but I don't know what to make of the display (see below).

1) Is ANOVA the right test? If not, can you tell me what is?

2) IF ANOVA is the right test, can you make some sense of the display?

. anova believe_alc_heart1 ev1_age

                           Number of obs =    4176     R-squared     =  0.0163                            Root MSE      = .789973     Adj R-squared = -0.0004

                  Source |  Partial SS         df       MS                 F       Prob > F               -----------+----------------------------Model |  42.5736723    70  .608195318       0.97     0.5380                          |              ev1_age |  42.5736723    70  .608195318       0.97     0.5380                          |              Residual |  2561.75367  4105  .624056924                

                     Total |  2604.32735  4175  .623790981

Re: Means between 3 or more groups

by Michael Peters -

Isaac,

 

  The ANOVA is the correct test, but this is only going to tell you if there is a difference in mean values across your three groups.  Generally, what we are interested in, is what is the difference between the groups themselves.  To do this properly you run an ANOVA and then correct the p-values for multiple comparisons.  The correction can be done a variety of ways, but the most common is the bonferroni or the sidak.  In addition, you usually want a summary of the mean values between groups.  Here is the command for these.

 

oneway continuousvariable catvariable, bonferroni tab

 

The readout will then give you a table of the means, an analysis for the overall difference, and then finally the total difference and the p-values for each comparison. If you would prefer the sidak correction over the bonferroni just replace bonferroni with sidak. Let me know if you have any other problems.

 

Michael