I am running a paneled and grouped boxplot and want to summarize the mean and standard deviation as a xaxis table below the plot. I have the below code currently:   proc sgpanel data=data; panelby panel/ layout=columnlattice onepanel uniscale=row novarname; vbox var/ category=location group=type nooutliers notches ; rowaxis values=(0 to 18 by 2) grid; colaxis display=(nolabel); colaxistable meanvar/ label="Mean" stat=median class=type classdisplay=cluster valueattrs=(size=5) ; keylegend / title=""; run;   Basically, when I try to run the colaxistable with var as the variable and stat=mean, it is putting too many decimals, so instead I created a separate table  with the mean and standard deviation (the standard deviation option is not available for colaxistable anyway) for each location, type group and merged it back in to the original data. Since I have multiple observations for var for each panel-location-type, I think it is giving me a sum in the graph, so I put stat=median as a workaround to give me the actual value I calculated but its still summing them. The other issue is the group variable, I want the value to be under the boxplot for the correct panel-category-group and it wasnt doing that, so I put the group variable as the colaxistable class and groupdisplay=cluster, but the order is wrong. Anyone have any tips or workaround? ... View more

I have some data where I have 2 dependent variables, method (1 or 2), and location (1, 2, 3, or 4). I ran an anova and got a significant result. Thus I want to do post hoc comparisons. I'm not interested in all comparisons, but only for location1 (method 1 vs 2), for location2 (method 1 vs 2), etc. How can I do this in proc glm?   I need these adjusted p-values to report in my paper. Is the only way to do it using contrasts and then adjust my alpha (bonferroni)? I think it will be too conservative. ... View more

I have some data like the below:   Group Sort Value  1 1 50 1 2 55 1 3 60 1 4 70 2 1 40 2 2 60 2 3 40 2 4 90   I want to create 2 new variables x and y such that if sort=1 then x=0 and y=value, otherwise x=value from previous row y and y= x + value, so like below:   Group Sort Value x y 1 1 50 0 50 1 2 55 50 105 1 3 60 105 165 1 4 70 165 135 2 1 40 0 40 2 2 60 40 100 2 3 40 100 140 ..   Can anyone help with code. Whenever I run using lag and by sort and group, the new y does update based on the previous calculation.  ... View more

Hello,   I have some data for a binary Y (Y/N) and a continuous X. I cut the X variable into categories based on increments of 10 (5 groups) and plotted the predicted probability. The relationship does not look linear to me. However, I would prefer to use the continuous version of the variable as I am interested in its relationship with Y. More specifically, I want to know the optimal cutpoint to discriminate Y.  I hypothesize that the association with the log odds of Y will increase and then be constant. I want to know that value of X that best predicts success, and after that higher X is no longer beneficial. What is the best way to do this? I am investigating adding an effect for a spline, but how do I know where the knots should be, as  I assume the knots are those critical values? Then, how can I estimate the OR for those intervals between the knots? ... View more

I have some data that looks like the below:   ID  TYPE 1     A 1     A 1     B 1     B 1     A     * 1     A     * 1     A     * 1     C 1     C 2     A ...   This data is already sorted by  a 3rd variable TIME, which is not shown here. Basically I was to flag when the count of a particular combination of ID and TYPE is greater than 2 (see starred rows). It is possible for example that 1A is followed by 1B but the count of each is less than or equal to two, I wouldnt flag those. I only want to flag it if they occur more than twice consecutively in the sequence it is already sorted by.    ... View more

I have some pre-post data on 2 groups. The data is binary (Y/N) and I am specifically interested in comparing the rate reduction between the 2 groups (the number of Yes's that turned to No after the intervention). What statistical test should I use to determine whether there is a statistically significant difference between the 2 independent groups?  ... View more

I want to use the nliteral function to rename a variable into the label, however, I need to parse the label. Is it possible to combine the nliteral with the scan function, or if not is there a workaround?  ... View more

I have the following date rows:   VAR1 This is row 1 (row1) This is row 2 (I think) (row2) (row3.)   And I need it to look like this: row1 row2 row3   Basically I need to trim the string into the text in between the last parenthesis. Can anyone help with code to perform this?  ... View more