1
$\begingroup$

I want to compare 2 values to see if there considerable significant different between them.

Basically these values comes from 2 credits score bureau providers (score 1 and score 2). The score 3 is just the avg between score 1 and 2. and we base our decision on this score 3.

+---------+-----------+----------+----------+
+       id+    score 1+   score 2+   score 3+
+---------+-----------+----------+----------+
+  006abXi+        100+        50+        75+
+---------+-----------+----------+----------+
+  006fsRe+        600+       550+       575+
+---------+-----------+----------+----------+
+  006csXm+        320+       350+       335+
+---------+-----------+----------+----------+

We are trying to identify if it is worth to keep using score 2, and just based our decision on the first credit bureau.

My current sample is a bit under 3500 records. And to have a very rough idea, I did an average of the score 1 and score 2 and see if there is much difference between them

+---------+-----------+----------+----------+
+      avg+    score 1+   score 2+   score 3+
+---------+-----------+----------+----------+
+         +     556.92+    584.14+    570.53+
+---------+-----------+----------+----------+

Doesn't look much but I do not think this will be the right way to base this decision on. I am wonder if there is any formula or sort of graph that I can use to visualize and be able to base my decision in some more scientific background?

$\endgroup$

1 Answer 1

1
$\begingroup$

You want to "see if there is considerable significant different between them".

With a sample size of 3500, your difference will very likely be significant. You could check using some appropriate paired difference test like the paired t-test or Wilcoxon signed rank test.

Whether the difference is considerable depends on what you want to do with it, what constraints exist in your task, and what you are willing to tolerate.

As far as visualization is concerned, I would suggest a histogram for the score differences.

$\endgroup$
2
  • $\begingroup$ Thank you yeah i was working on histogram and did a few charts of how many people are between certain ranges, that helped a lot, I will check out the rank test $\endgroup$
    – Manza
    Jul 28, 2022 at 5:58
  • $\begingroup$ With large datasets, the significance tests usually always state significance, so it is more important to look at the effect. Visualization is a good tool in this regard. $\endgroup$
    – frank
    Jul 28, 2022 at 6:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.