Full A/B Test From Raw Counts

A different Marlow & Finch experiment, run in the autumn on the product page: Variant A shows the delivery charge only at checkout, Variant B shows it under the price straight away. The worry was that being honest earlier would scare people off; the hope was that it would stop them abandoning the basket later. Nobody on the team would confidently predict the direction, which is precisely why it was tested.

This one is end to end. You are given raw counts, with no rates worked out for you this time, and asked for a verdict, with every intermediate step left to you. The same script is pointed at every experiment in the archive, and those arms are not always the same size.

The conventions, which are the same ones the earlier problems in this set used:

• The test is on proportions — the share of product-page visits that ended in an order — not on averages of amounts.
• Judge it against the null hypothesis that the two pages perform identically, and size the expected wobble under that assumption, treating both arms as sharing one rate.
• Two-tailed, because a drop would have been just as important as a rise. Only the distance from zero counts.
• The critical value comes in with the experiment; the boundary itself does not count as significant. At the 5% level it is 1.96.

Task: Print exactly significant if the difference between the two arms clears that bar, and exactly not significant if it does not. No quotation marks, nothing else on the line.

Input

Five lines, each holding one number, in this order: the control arm's orders, the control arm's visitors, the challenger arm's orders, the challenger arm's visitors, and the critical value. The first four are whole numbers and each visitor count is at least 1; the critical value is a positive decimal.

Output

One line holding one of the two phrases above, in lower case, and nothing else.

Example:

Input:
120
2000
150
2000
1.96

Output:
not significant