Statistical Comparison of Product: High Average vs. Low Range

FPT2001

Inactive Registered Visitor
#1
I am looking for suggestions on how to statistically compare products, not processes.

We manufacture springs that are the same length but have different widths and thicknesses. We are evaluating as-is and 1, 2, 3 hour heat treat. We want to compare these products with competitor products. There is only 1 finished specification/attribute: resistance.

I am looking for help writing a protocol to compare our different product to competitors. Initial testing shows some products have larger Std. and Ranges than the others but the Mean is much higher.

How do I evaluate benefits of a higher Mean compared to other products that have a much tighter range and/or Stddev?
 

Bev D

Heretical Statistician
Staff member
Super Moderator
#2
Re: Comparison of Product: High Avg vs Low Range

I think you may have the statistics cart in front of the physics horse.

A useful statistical analysis can only be determined once the real question is answered: How are your parts used and what is important to your customers?

Without physics, statistics is just gambling.
 

FPT2001

Inactive Registered Visitor
#4
Horse and cart are just fine. We know our customers needs and wants well. You have apparently never worked with sales and marketing.

Given a customer requirement that does not identify number of cycles to failure or minimum cycles to failure... we are evaluating competitor products as well as several versions of a new product we want to introduce.

We are looking for the best statistical methods for comparing product that have a higher cycle time/high Std Dev with lower cycle time/lower std dev. Something along the lines with coefficient of variation.
 

Miner

Forum Moderator
Staff member
Super Moderator
#5
Bev meant no disrespect, but you have not given us much information to go on. In your first post, I thought you were talking about the spring rate. In your second post, it sounds like you are talking about the expected life, cycles to failure.

Please give us more detail, and we will try to help.
 

FPT2001

Inactive Registered Visitor
#6
Its impossible to determine tone of a text message, I didnt take it as disrespect and my reply wasnt negative. Technology still hasnt duplicated the humor/human element into a text message.

Let me just start over and simplify my question.

I have a product that does not have a Min or Max specification. The products are tested for time to failure. We want to market a new product that has a greater time to failure. When we tested our product against competitors we found we had a higher time to failure but the range and Std Dev was much higher.

We can look at the Cv for each (std dev/Mean) and make an assumption concerning the variation. However, we are looking for something to help with the marketing and sales promotion (easier to quickly understand). For instance, how can you say Product E performs better than the others even though the Std Dev and Range are higher?

Product A B C D E
Average 90 400 140 110 530
STD Dev 13 47 30 25 112
Range 38 120 100 102 385


I do appreciate your help.
 

Jim Wynne

Super Moderator
#7
Its impossible to determine tone of a text message, I didnt take it as disrespect and my reply wasnt negative. Technology still hasnt duplicated the humor/human element into a text message.

Let me just start over and simplify my question.

I have a product that does not have a Min or Max specification. The products are tested for time to failure. We want to market a new product that has a greater time to failure. When we tested our product against competitors we found we had a higher time to failure but the range and Std Dev was much higher.

We can look at the Cv for each (std dev/Mean) and make an assumption concerning the variation. However, we are looking for something to help with the marketing and sales promotion (easier to quickly understand). For instance, how can you say Product E performs better than the others even though the Std Dev and Range are higher?

Product A B C D E
Average 90 400 140 110 530
STD Dev 13 47 30 25 112
Range 38 120 100 102 385


I do appreciate your help.
I'm not sure whether you're looking for a way to analyze data or a way to spin the data you already have. If I understand correctly, what you want to know is given identical load characteristics, which spring (yours or your competitor's) is most likely to fail first. Is this correct?
 

Miner

Forum Moderator
Staff member
Super Moderator
#9
First, be sure that your data is normally distributed. While not unheard of, it is unusual for reliability data to be normally distributed. Then I would determine and compare the B10 lives of the products. B10 is the tenth percentile, or the time at which 10% of the product could be expected to have failed.
 

reynald

Quite Involved in Discussions
#10
How do I evaluate benefits of a higher Mean compared to other products that have a much tighter range and/or Stddev?
Are you talking about the mean/std dev of the resistance here? i have the impression that this can be done by showing the histogram of the resistances and adding the specifications as a reference. Then compute the process capability. By then you could objectively assess which one is better, the higher mean or the tighter range. I guess that should do it unless i totally mis-interpreted the question.
Regards,
Reynald
 

Top Bottom