more than 2 groups found, only 2 allowed r(499)

[Nida Latif]

Hello,

I'm trying to run a mann whitney u test to compare the medians between cfrnonLAD and cfrLAD

but I keep getting this error message:
. ranksum cfrnonlad, by(cfrlad) porder exact
more than 2 groups found, only 2 allowed
r(499);

any idea how to fix this? I'm a complete newbie to stata.

Thanks!

[Joseph Coveney]

. . . any idea how to fix this?

You could try something like the following.

rename cf*LAD lad#, addnumber(0)
generate long row = _n
reshape long lad, i(row) j(grp)
ranksum lad, by(grp)

[Joseph Coveney]

You can use stack as an alternative.

stack cfrnonLAD cfrLAD, into(lad)
median lad, by(_stack)

[Nida Latif]

Thank you, it worked when I used the code:
rename cf*LAD lad#, addnumber(0) generate long row = _n reshape long lad, i(row) j(grp) ranksum lad, by(grp) However, when I used the following code:

stack cfrnonLAD cfrLAD, into(lad)
median lad, by(_stack)

It gave me the results of pearson's chi square instead of mann whitney u test.

[Joseph Coveney]

median lad, by(_stack)

. . . instead of mann whitney u test.

Yes. In your original post you said that your objective is "to compare the medians", and the Mann-Whitney U test isn't strictly a test of the equality of medians.

[Nida Latif]

Thank you. This is really helpful!

[Joseph Coveney]

Thank you.

You're welcome. Thank you for closing out the thread.

[Nick Cox]

Always plot the data too. Even with two groups, the difference can be much more complicated than just a shift, as shown by a difference in medians.

https://www.statalist.org/forums/for...oup-histograms gives examples of side-by-side comparison focused on medians and quartiles AND means AND indeed all the detail in distributions.

[Nida Latif]

Hello,

I tried the same code to run the mann whitney u test to compare the median of baselinelessthan50 and baseline greaterthan50

rename baselin*greaterthan50 greaterthan50#, addnumber(0)
generate long row = _n
reshape long greaterthan50, i(row) j(grp)
ranksum greaterthan50, by(grp)

but its giving me the following error:

any idea how to fix this?

[Nick Cox]

Your reshape just left the dataset in the same shape with different variable names. It would be easier to help if you gave us a data example of the original data (not a screenshot).

Please read and act on

https://www.statalist.org/forums/help#stata

and follow up with

dataex, count(160)

[Joseph Coveney]

I tried . . . the mann whitney u test to . . . baselinelessthan50 and baseline greaterthan50
. . . any idea how to fix this?

It'll probably be more straightforward to use stack for this.

stack *than50, into(outcome) clear
tabulate _stack
assert r(r) ==  2
ranksum outcome, by(_stack)

[Nida Latif]

Thank you so much Nick and Joseph for all your help. I really appreciate it! It worked when I used stack like Joseph suggested.