Noob Needs Help! Please Please Please Help!

Question

MD doing reserach on Genetic Mutations, I really need help on this, I have never used PL/SQL. Can somebody help me with this?
I have a view made of mulitple table that has following columns: diagnosis, patient_id, kindred_id, column_d, colum_e.... column_p where each distinct combination of column_d to column_p means one specific mutation. I want to group patients who have same mutation to find out the following:
for each mutation(unique column d through column p), I want to find out how many families have diagnosis 1, diagnosis 2, diagnosis 3...
The rule is the program will look at kindre_id first, to see if this person's kindre_id has been looked at. If
so, jump it/ignore it. If this kindred_id has never been looked at, treat it as a unique family. Diagnosis count for that diagnosis will then increase by 1. Sounds easy, but I don't know how to write that for loop that will do for each unique (column d, column c, ... column p) do the following... Does such a thing exist?

Answer 1

Hi,
SELECT     diagnosis
,     COUNT (DISTINCT kindred_id)     AS family_cnt
,     column_d
,     column_e
,     column_f
FROM     table_x
GROUP BY     diagnosis
,          column_d
,          column_e
,          column_f
;     produces this output:
DIAGNOSIS FAMILY_CNT COLUMN_D   COLUMN_E   COLUMN_F
         1          2 abc        bca        acb
         2          2 abc        bca        acb
         3          2 abc        bca        acbwhich is the data you want, but not in the format you want. You want to pivot the family_cnt column from the three rows into three columns in one row. For general information about how to this, search for "pivot" on this or similar sites.
One solution is:
SELECT     COUNT (DISTINCT CASE WHEN diagnosis = 1 THEN kindred_id END)     AS diagnosis_1
,     COUNT (DISTINCT CASE WHEN diagnosis = 2 THEN kindred_id END)     AS diagnosis_2
,     COUNT (DISTINCT CASE WHEN diagnosis = 3 THEN kindred_id END)     AS diagnosis_3
,     column_d
,     column_e
,     column_f
FROM     table_x
GROUP BY     column_d
,          column_e
,          column_f
;     which produces the following output:
DIAGNOSIS_1 DIAGNOSIS_2 DIAGNOSIS_3 COLUMN_D   COLUMN_E   COLUMN_F
          2           2           2 abc        bca        acbThis technique assumes that you know exactly how many different diagnoses there are and what the distinct values of the diagnosis column are. (I.e., in this example, I knew that that there were three diagnosie and that their values were 1, 2 and 3. I used that information to write three COUNT expressions, with values 1, 2 and 3 hard-coded into them.)
So, what can you do if you do not know the diagnosis values? The columns have to be hard-coded into the query, so how can you possiblly write something today that will have all the values that are in the table next month? You can't, but you can write something today that you can run (unchanged) next month which will write such a query. In other words, you can do dynamic SQL, where you run a preliminary query to get the different diagnosis values, write an appropriate query similar to the one above, and then run that newly-created query. The PL/SQL command "EXECUTE IMMEDIATE" can help with this.
The code above works in all versions of Oracle. In Oracle 11, there is a more straigtforward way of specifyng PIVOT in a SELECT statement.