It sounds like what you want is a plot of the spline fit with the original variable on the horizontal axis, the vertical axis is the hazard ratio, and the points on the plot are the hazard ratios for changing the original variable from a reference value (like 20 as you mention) various amounts. That can be done as has already been shown using the HAZARDRATIO statement with the UNITS= option. You can then save the resulting table of hazard ratios and then use PROC SGPLOT to plot it. For example, using the myeloma data in an example in the PHREG documentation: proc phreg data=Myeloma; effect spl=spline(protein/naturalcubic); model Time*VStatus(0)=spl ; hazardratio protein / at (protein=10) units=1 to 5; ods output hazardratios=hr; run; proc sgplot noautolegend; highlow high=waldupper low=waldlower x=description / lowcap=serif highcap=serif; scatter y=hazardratio x=description; xaxis label='Protein' grid; yaxis label='Hazard Ratio' grid; refline 1 / axis=y; run; ... View more

In addition to getting the statistics as I just showed, if you also want the AUC, there is an analog of this for the multinomial case that you can obtain using the MultAUC macro that is available in this note. ... View more

Okay, but rather than what you've suggested, consider fitting the GLOGIT model to your nominal, multinomial  response and producing the two-way table of actual by predicted values. You can then simply collapse rows and columns of the table as needed for each level to produce the usual statistics you can get with a 2x2 table.   See the example using the Crops data in this note which fits the GLOGIT model and obtains the actual by predicted table. The statements below repeat that and then computes, for the Clover level, the sensitivity (recall), specificity, positive predictive value (precision), and negative predictive value using the SENSPEC option in PROC FREQ. You can do similarly for each of the other Crop levels. Using this approach, you could compute other statistics, like accuracy, as shown in this note.   The _INTO_ variable in the PREDS data set contains the predicted Crop values from the fitted model. The first FREQ step shows the full predicted by actual table. PROC FORMAT creates a format that groups all of the levels that are not Clover together so that the second FREQ step with the FORMAT statement collapses the table into a 2x2 table for Clover vs not Clover and produces the statistics.  proc logistic data=Crops; model Crop=x1-x4 / link=glogit; output out=preds predprobs=individual; run; proc freq data=preds; table _INTO_*crop / out=CellCounts; run; proc format; value $notcl 'Clover'='Clover' other='NotClover'; run; proc freq data=preds; format _INTO_ crop $notcl.; table _INTO_*crop / senspec; run; ... View more

You'll need to clarify your goal. What do you mean by "finding the one-vs-the-rest power of a single species" expressed in terms of a statistical model or test? If you want to test IRIS vs the others with respect to the multivariate means across the 4 variables, then an example of exactly that is shown in the example titled "Multivariate Analysis of Variance" in the PROC GLM documentation. Note that this reverses the model by making SPECIES the predictor of the 4 variables. Note that you could write a single model with SPECIES as the response variable to deal with all of the class levels simultaneously (model species=SepalWidth SepalLength PetalWidth PetalLength / link=glogit;) rather than create binary response variables and do separate analyses. And you could get predicted probabilities from this single model for each observation being in each of the species (output out=preds predprobs=individual;), but what exactly would you want to test with this? ... View more

When you use PROC GLMSELECT (like PROC GLM or PROC REG) you are assuming that the response is approximately normally distributed. If you have a response which is distributed otherwise, such as if your response is a count, is categorical, or is positively-valued and skewed, and you want to use LASSO selection, then you can use PROC HPGENSELECT and specify a suitable response distribution with the DIST= option. ... View more

Yes, sorry, I had just been using 1 initially. ... View more

If observations are matched then they are not independent. Study data can consist of matched sets of any size, not just pairs, with any number of events observed within each set if the response is binary. There are various analytical methods that can be used for binary response data of this type including conditional logistic regression which is available with the STRATA statement in PROC LOGISTIC. Other possibilities include the Generalized Estimating Equations and Alternating Logistic Regressions using the REPEATED statement in PROC GEE or PROC GENMOD and random effect models using the RANDOM statement in PROC GLIMMIX. A non-model-based approach using stratification is available with the CMH option in PROC FREQ when a multi-way table is specified. See the examples in the documentation for each of these procedures. ... View more

For the case of the lognormal, if you use FMM, the predicted values that it generates are on the original scale, but in GLIMMIX they are on the log transformed (normal) scale even though both use the identity link as has been noted. This distinction is pointed out in the description of the lognormal log likelihood in the FMM documentation.   Here is an example.  Again as has been pointed out, if you can use an alternative distribution like gamma or inverse gaussian with a suitable link function, that will avoid the problem by using ILINK in LSMEANS or when getting predicted values. data L; do x=1 to 10; do n=1 to 1000; y=rand('lognormal',x); logy=log(y); output; end; end; run; proc means data=l; class x; var y logy; run; proc fmm data=l; class x; model y=x/dist=lognormal cl; output out=l2 pred=p; run; proc means data=l2; class x; var p ; run; proc glimmix data=l; class x; model y=x/dist=lognormal cl s; lsmeans x/ilink; output out=l3 pred(ilink)=p; run; proc means data=l3; class x; var p ; run; proc glimmix data=l; class x; model y=x/dist=gamma cl s; lsmeans x/ilink; output out=l4 pred(ilink)=p; run; proc means data=l4; class x; var p ; run;   ... View more

If you find evidence of overdispersion, either from a too large Value/DF or from the test using SCALE=0 and NOSCALE with DIST=NEGBIN, then that suggests that the distribution is not Poisson and something else should be used like in the note I referred to. The warnings don't play into it. There is no hard and fast cutoff criterion on the Value/DF.  ... View more

You didn't provide any information about the nature of your response variable, but it sounds like it might be positively- or at least nonnegatively-valued. Such responses could be modeled using a generalized linear model using various distributions. If it is a discrete count response, then you could use the Poisson distribution (or negative binomial if it exhibits extra variability or excessive zeros). If it is continuous and positive, then the gamma, inverse Gaussian, or Tweedie distributions are commonly used. With these distributions, it is common to use the log link function to model the log of the distribution mean.   But yes, if your subjects or items are repeatedly measured, then the implied correlation should be taken into account by using an appropriate model such as a Generalized Estimating Equations model (PROC GEE) or a random effects model (PROC GLIMMIX) depending on the purpose of the model. Both procedures can use distributions like Poisson and gamma. See the examples in the documentation of those procedures. Also see this note on modeling positive, continuous responses and this note on modeling count responses using various distributions (though this is primarily about handling overdispersion). ... View more