Bar graph not showing correct mean for one variable

[Viktor Elmberg]

Hi!

I'm trying to create a bar graph with CI's for a total of 6 variables using catcibar with the following code:

catcibar MDP_A1 MDP_SQ_1int MDP_SQ_2int MDP_SQ_3int MDP_SQ_4int MDP_SQ_5int  if once == 1, over(lastb1) color (eltgreen erose) ytitle("NRS 0-10") xtitle ("MDP Dimension")

All variables are shown in the bar graph with correct means except MDP_SQ_1int which is shown as having a too low mean value which I can also confirm by running a ttest or mean command. Confusingly, I seem to get the correct values for MDP_SQ_1int if run catcibar with only this variable. Any suggestions on how I could resolve this would be greatly appreciated.

[Nick Cox]

Please see the FAQ Advice especially https://www.statalist.org/forums/help#stata for good practice in asking questions.

You don't explain catcibar but a Google reveals

 
 net install catcibar, from("https://aarondwolf.github.io/catcibar")

So far, so good, but I can only make guesses in the absence of a data example such as we request.

I suspect that you have problems with missing values. I would ideally like to see the results of

 
 list lastb1 MDP_A1 MDP_SQ_1int MDP_SQ_2int MDP_SQ_3int MDP_SQ_4int MDP_SQ_5int  if once == 1

if that's not too voluminous and minimally the results of

  
 su lastb1 MDP_A1 MDP_SQ_1int MDP_SQ_2int MDP_SQ_3int MDP_SQ_4int MDP_SQ_5int  if once == 1

[Viktor Elmberg]

Thank you, I will try to explain the question better next time.
I discovered that the last variable (MDP_SQ_5int) also gave a slightly wrong mean using catcibar.
All missing is marked with ".".
The first option was indeed very voluminous. The results of the second is:

Variable	Obs	Mean	Std. dev.	Min	Max
lastb1	92	1.217391	.4147311	1	2
MDP_A1	88	4.261364	2.146873	0	8
MDP_SQ_1int	86	4.011628	2.596351	0	10
MDP_SQ_2int	81	1.320988	2.096587	0	9
MDP_SQ_3int	82	.9756098	1.609966	0	8
MDP_SQ_4int	83	2.168675	2.453565	0	8
MDP_SQ_5int	89	4.696629	2.785533	0	10

[Nick Cox]

There may be other problems -- but you should check out the cw option.

cw                      Calculate means and confidence intervals casewise. Default is to drop all observations where any observations of varlist, over, or
                                by(varlist) are missing.

[Viktor Elmberg]

Thank you so much. Been working on this for hours and this did the trick instantly.