spearman rank correlaiton coeffiecent on excel?

  • Thread starter Thread starter Guest
  • Start date Start date
tony -

If you browse to google.com and search for

spearman rank correlation coefficient on excel?

you'll get 38,300 hits. Have you tried that? (It helps if you spell
"correlation coefficient" correctly.)

- Mike
www.mikemiddleton.com
 
The answer according to Sussex uin is no excel dosent calculate Spearman,
however their site, to which I was directed to on your advice expains how I
can manipulate excel to perform the calculation; so thank you
 
tony said:
The answer according to Sussex uin is no excel dosent
calculate Spearman,
however their site, to which I was directed to on your advice
expains how I
can manipulate excel to perform the calculation; so thank you
Click to expand...

I wonder why you would actually want to use the rank correlation
method? I always thought it was a remnant of the days when more
exact calculation was tedious but I'd be glad to be enlightened.
 
James -
I wonder why you would actually want to use the rank correlation method? <
Click to expand...

Spearman's correlation is most appropriate when you want a correlation
measure for two variables that are ordinal categorical measures (instead of
numerical measures). For example, it may not make sense to arbitrarily
assign numerical values (1,2,3,4,5) to ordinal responses on a survey
questionnaire (Strongly Disagree, Disagree, Indifferent, Agree, Strongly
Agree)
I always thought it was a remnant of the days when more exact calculation
was tedious but I'd be glad to be enlightened. <
Click to expand...

Pearson's correlation (Excel's CORREL worksheet function) summarizes a
linear relationship, so Spearman's correlation could be used to summarize a
nonlinear relationship between two numerical variables. Also, Spearman's is
not influenced as much by outliers. But, if you truly have a linear
relationship between two numerical variables, you lose information if you
convert the numbers to ranks before computing correlation.

- Mike
www.mikemiddleton.com
 
Mike wrote on Mon, 8 May 2006 19:43:45 -0700:

??>> I wonder why you would actually want to use the rank
??>> correlation method? <

MM> Spearman's correlation is most appropriate when you want a
MM> correlation measure for two variables that are ordinal
MM> categorical measures (instead of numerical measures). For
MM> example, it may not make sense to arbitrarily assign
MM> numerical values (1,2,3,4,5) to ordinal responses on a
MM> survey questionnaire (Strongly Disagree, Disagree,
MM> Indifferent, Agree, Strongly Agree)

??>> I always thought it was a remnant of the days when more
??>> exact calculation was tedious but I'd be glad to be
??>> enlightened. <

MM> Pearson's correlation (Excel's CORREL worksheet function)
MM> summarizes a linear relationship, so Spearman's correlation
MM> could be used to summarize a nonlinear relationship between
MM> two numerical variables. Also, Spearman's is not influenced
MM> as much by outliers. But, if you truly have a linear
MM> relationship between two numerical variables, you lose
MM> information if you convert the numbers to ranks before
MM> computing correlation.

Thanks Mike, that does put things in a better perspective
especially non-linearity and the effect of outliers.


James Silverton.
 
Back
Top