topic Re: PROC LOGISTIC - Journal Note "There are no valid observations" but code still runs in New SAS User

PROC LOGISTIC - Journal Note "There are no valid observations" but code still runs

LG-ins-2022 — Thu, 01 Dec 2022 22:10:36 GMT

Hi SAS experts,

So I am scratching my head on this one. I am running a proc logistic with several variables, for 2 of them (who each have less than <5% of values missing), the SAS journal writes :

NOTE: There are no valid observations for independant var.

I recoded for each variable in the model the missing values.

I also tried a proc freq for each independant var*dependant var, without anything weird.

As stated in the title, I still have an output (and the OR for each categories).

Can someone help me understand this note?

Cheers

Re: PROC LOGISTIC - Journal Note "There are no valid observations" but code still runs

PaigeMiller — Thu, 01 Dec 2022 20:03:48 GMT

NOTE: There are no valid observations for independant var.

Doesn't this lead you to the problem directly?

If not: without seeing the log and parts of your data, its really hard to provide advice.

Please show us the ENTIRE log for PROC LOGISTIC (not selected parts, not just the errors or warnings, but every single line of the log for PROC LOGISTIC) by copying the log as text and pasting it into the window that appears when you click on the </> icon.

We would also need to see a portion of your data, including the independent variable.

Re: PROC LOGISTIC - Journal Note "There are no valid observations" but code still runs

LG-ins-2022 — Thu, 01 Dec 2022 22:12:17 GMT

14799  Proc logistic data=data;
14800      weight weight;
14801      class
14802
14803      Ct12cM (ref="Tout à fait ou plutôt satisfait" missing)
14804      Ct22cM (ref="Tout à fait ou plutôt satisfait" missing)
14805      Ct32cM (ref="Tout à fait ou plutôt satisfait" missing)
14806
14807      genre_3C (ref="Homme" missing)
14808      age_repondant3cM_EV (ref="65 ans et plus" missing)
14809
14810      ScolariteM_EV (ref ="pas de diplôme" missing)
14811
14812      Revenu4cM_EV (ref=last missing)
14813
14814
14815  /param=glm;
14816      model risque (event = "Danger") =
14817      Ct12cM Ct22cM Ct32cM
14818      genre_3C  age_repondant3cM_EV ScolariteM_EV
14819      Revenu4cM_EV
14820  ;
14821
14822  run;

NOTE: PROC LOGISTIC is modeling the probability that risque ='Danger'.
NOTE: Convergence criterion (GCONV=1E-8) satisfied.
NOTE: There are no valid observations for age_repondant3cM_EV=Âge manquant.
NOTE: There are no valid observations for ScolariteM_EV=Scolarité manquantes.
NOTE: There were 10036 observations read from the data set WORK.CHALEUR10.
NOTE: PROCEDURE LOGISTIC used (Total process time):
      real time           0.31 seconds
      cpu time            0.31 seconds

Good afternoon,

Please find the log attached.

Cheers

Re: PROC LOGISTIC - Journal Note "There are no valid observations" but code still runs

PaigeMiller — Thu, 01 Dec 2022 20:06:53 GMT

Repeating with emphasis: Please show us the ENTIRE log for PROC LOGISTIC (not selected parts, not just the errors or warnings, but every single line of the log for PROC LOGISTIC)

Re: PROC LOGISTIC - Journal Note "There are no valid observations" but code still runs

LG-ins-2022 — Thu, 01 Dec 2022 20:12:35 GMT

Please see my answer edited

Re: PROC LOGISTIC - Journal Note "There are no valid observations" but code still runs

PaigeMiller — Thu, 01 Dec 2022 20:33:14 GMT

So, even in your initial post, you truncated the NOTE in the log, altering its meaning. Please don't do that.

NOTE: There are no valid observations for age_repondant3cM_EV=Âge manquant.
NOTE: There are no valid observations for ScolariteM_EV=Scolarité manquantes.

There are no valid observations for variable age_repondant3cM_EV when it equals Âge manquant. So look at your data when this variable equals Âge manquant and see if there are valid observations (possibly the response variables are all missing, possibly other predictor variables are all missing). Same thing for the other variable listed.

Re: PROC LOGISTIC - Journal Note "There are no valid observations" but code still runs

LG-ins-2022 — Thu, 01 Dec 2022 22:13:10 GMT

Duly noted!

As I mentioned in my original post, both my proc freq tables indicates there are mostly valid observations for my dependant variable.

When I double-checked for both variable and my dependant variable, there is 1 event with both variables missing.

Could that be enough for the log to indicate such note?

Cheers

Re: PROC LOGISTIC - Journal Note "There are no valid observations" but code still runs

PaigeMiller — Thu, 01 Dec 2022 21:40:58 GMT

I specifically suggested you look at other predictor variables, and the response variable.