Yes. I think so. And better use numeric type variable. ... View more

@SAS-questioner wrote: 3. Since the polytomous outcome is a three-level categorical outcome, why in the odds ratio output, there is only one odds ratio for group and time each? In this case, how to interpret the 'median', and 'low' using odds ratio? For example, comparing group 1, group 0 was 3 times more likely be in the 'large'. But how to find out the odds ratio for 'median'/'low'? Home > Analytics > Stat Procs > Obtaining odds ratio estimates for each level of ordinal variable in PROC GLIMMIX https://communities.sas.com/t5/Statistical-Procedures/Obtaining-odds-ratio-estimates-for-each-level-of-ordinal/td-p/725370   BR, Koen ... View more

Some how you have to tell SAS that 3 was a possible answer. PROC SUMMARY will let you do that by attaching a FORMAT to the variable. proc format ; value cat 1='1' 2='2' 3='3' 4='4' ; run; proc summary data=have nway completetypes; class cat1 cat2 / preloadfmt; format cat1 cat2 cat.; output out=counts ; run; Result Obs cat1 cat2 _TYPE_ _FREQ_ 1 1 1 3 1 2 1 2 3 1 3 1 3 3 0 4 1 4 3 1 5 2 1 3 1 6 2 2 3 1 7 2 3 3 0 8 2 4 3 1 9 3 1 3 1 10 3 2 3 1 11 3 3 3 0 12 3 4 3 1 13 4 1 3 1 14 4 2 3 1 15 4 3 3 0 16 4 4 3 1 ... View more

As Freelance said, If you are looking for Kendall's W. Check %magree macro.     ... View more

Hello,   In your first PROC CATMOD, you might need the ZERO= option behind the forward slash on the MODEL statement. ZERO=keyword specifies whether a nonmissing cell with zero weight in the data set is treated as a sampling or structural zero.   You can also take a PROC FREQ approach (calculating agreement measures).   Koen ... View more

Question (actually a trick question) - how do you know that the distribution for residuals is not normal? Did you do some sort of test? There are well-known issues with almost every hypothesis test for normality (overpowered with N greater than about 40, underpowered for N less than about 15), and the linear mixed model is remarkably robust to the assumption of normality of the residuals, so long as the empirical distribution is mono-modal, not truncated, and lacks extremely large absolute values. The mono-modal basically boils down to sex differences.   So here are some ways to attack the issue, from simple to complex: bin your responses to four or five categories and consider using Cochran-Mantel-Haenszel methods where you stratify by sex. Plot your data and see what the shape looks like. From that, use a generalized linear model, assuming the distribution you have a picture of. If you have what might be considered random effects, use a generalized linear mixed model. Bootstrap your data. Simulate a lot of datasets that could possibly occur based on your current data. Use a Bayesian analysis with noninformative priors. This does a lot better job of simulating the data needed to construct credible intervals as you can include correlations over time or clusters. I don't think you have any random effects, so a good start on this can be found by looking through the documentation for PROC BGLIMM. Given what you have done so far, I would recommend #4. You can use most of your PROC MIXED code, and you can examine each distribution/link to see which best fits your data.   SteveDenham ... View more

@SAS-questioner wrote: ... stuff deleted ... How did you come up this idea? ... Forget the functions for generating N_GAPS.  The tool of primary interest here (the "idea") is the HISTORY array.   The idea of using an array indexed by dates is an old one, but (to my knowledge) is not a technique taught in SAS programming classes.  (Only so much can fit in the syllabus).    There are two important aspects of this array:    It is declared as _TEMPORARY_, so It does not create variables that will be in the new data set. Its values are automatically RETAINED from obs to obs, so you can build the array over a sequence of observations. The array has lower and upper bounds corresponding to the dates 01jan2020 and 31dec2022.   Regarding the lower/upper bounds point:  Usually you can declare an array like array history {1096} _temporary_; which has elements history{1} through history{1096}.   But you can also declare an array like this: array history {21915:23010} _temporary_ ; which also has 1096 elements, but the leftmost element is history{21915} (not history{1}) and the rightmost is history{23010} instead of history{1096}.   And why did I choose those values as the bounds?  Because they are the SAS date values for 01jan2020 and 31dec2022 (which I took as the likely earliest and latest dates in your study).   Unfortunately, SAS doesn't accept this intuitively obvious syntax  array history {"01jan2020"d:"31dec2022"d} _temporary_ ;   So, because I didn't want to manually determine the actual date values 21915 and 23010, I used the macro function %sysevalf to do it for me.  But the macro function is just icing on the cake.   ... View more

Yeah that explains, I tried to put it into different data step and now it works this time. ... View more

Steve, I think Poisson Distribution is used for low probability event,like insurance's claim probability. While Logistic Regression is used for normal probability(not too small), otherwise need EXACT logistic model or Penalty Logistic model . ... View more

Hi, Ballardw, thank you for replying. I will put the expected data below:   County #_store product_A product_B #_need_A #_need_B 1 50 20 30 100 120 2 40 10 20 80 90 3 55 25 35 150 160 "#_store" is the total number of store that I selected in that county. The "product_A" is the number of stores that have product A, the "product_B" is the number of stores that have product B, "#_need_A" is the number of people that need product A in that county, "#_need_B" is the number of people that need product B in that county. And I want to compare product_A/#_need_A vs. product_B/#_need_B.   For "find out how many people (with certain feature)" is defined as "#_need_A" and "#_need_B", but currently, I am not sure if the "#_need_A" and "#_need_B" is a absolute numbers or incidence rate, if it's incidence rate, for example, 100 out of 10000 people, do I still use 100? Or 100/10000 in this case?   For "I select some counties first, and select some stores under each county. ", it's "county" variable in that data, since each county has different population or number of stores, I am wondering if I need to make it as a random effect? Or it's just regular rate comparison? ... View more

OK, I will create an new one, thank you! ... View more

Work like charm! Thank you! ... View more