using the correct stat test....

hi all



i have two groups of people. the control group has n=9 and the patient group has n=27.



i have data of the ages of all subject and i want to test if there is a significant difference in the age of the patients.



what is the correct t-test to perform and why?



my initial thought was mann-whitney u (non parametric for different sample sizes) but i don´t really know....any help would be GREAT!

Hello

There are a number of ways to screenout a statistical test for analysing the data

Selecting a statistical test. For your purpose I believe Mann whitney will do fine.

Use the following link to determine the test.

http://www.whichtest.info/index.htm

Hi! thank you for your reply and the link to what appears to be a very useful page.

perhaps this is a silly follow up question, but would age not be consider "interval" data ?

as according to the website you directed me to..."An INTERVAL scale is one in which intervals at different points on the scale are equal. Examples are the Celsius and Fahrenheit temperature scales. The difference between 20 and 22 degrees is the same as the difference between 15 and 17 degrees."

Thus, following this, the test recommened would be a Independent samples t-test if my data would be normally distributed....which is the next thing!!

i´ll put my data here cause its pretty straight forward.....

Age of controls....

78

66

67

64

79

62

50

44

61

Patients....

66

86

77

78

67

73

84

78

61

56

86

58

76

49

71

70

64

49

80

64

69

80

83

70

89

64

60

80

65

Any more help greatly appreciated!!!!!

http://www.minitab.com/en-US/training/tutorials/method-chooser.aspx

You may also find this basic guide to statistics (published by minitab) to be of use to you.  I have used it a few times in the past.

Yes I think you are correct to say this is data on an interval scale.

You can test whether or not your data set follows a normal distribution using the The Shapiro-Wilk Normality Test.

This tests whether values (X_i for i = 1 to n ) follows a normal distribution.

FYI, both your controls and patent data are Normal at the 0.05 level

Now when you do you your independent t-Test on the two smaples you need to ask yourself what you are testing.

1) Mean Ctrl - Mean Patents <> 0 ?  i.e. Are the mean ages of the two groups different?

2) Mean Ctrl - Mean Patients > 0 ?  i.e. Is the mean age of the Patient group younger than the mean of the Ctrl group?

3) Mean Ctrl - Mean Patients < 0 ? i.e. Is the mean age of the Patient group older than the mean of the Ctrl group?

 and of course set your significance level. Each test may have a different outcome as you can see below.

so for #1

Null Hypothesis:                      Mean1 - Mean2   =   0

Alternative Hypothesis:               Mean1 - Mean2  <>   0

                                                 

    t              DoF            P Value        

    ------------------------------------------------

    -1.75305       36             0.08811        

    ------------------------------------------------

                                                 

    At the 0.05 level, the difference of the population means                              

    is not significantly different than the test difference (0).    

for #2

Null Hypothesis:                      Mean1 - Mean2  <=   0

Alternative Hypothesis:               Mean1 - Mean2   >   0

                                                 

    t              DoF            P Value        

    ------------------------------------------------

    -1.75305       36             0.95595        

    ------------------------------------------------

                                                 

    At the 0.05 level, the difference of the population means                              

    is not significantly greater than the test difference (0).

For #3

Null Hypothesis:                      Mean1 - Mean2  >=   0

Alternative Hypothesis:               Mean1 - Mean2   <   0

                                                 

    t              DoF            P Value        

    ------------------------------------------------

    -1.75305       36             0.04405        

    ------------------------------------------------

                                                 

    At the 0.05 level, the difference of the population means                              

    is significantly less than the test difference (0).

Dear Fraser

Fantastic explanation. Thanks very much for your help. I truly appreciate it.

 

Best regards

David