在C＃中，应使用哪种数据类型存储具有标准偏差的数字列表？

Question

I'm performing some tests and I have a list of values with errors of the form:

12.7 ± 0.3
14.2 ± 0.1
70.8 ± 0.5

I need to keep the standard deviation alongside the value, as I need to use it for various calculations later on.

At the moment I'm just using

List<KeyValuePair<double, double>>

But are there any better solution?

Answer 1

If the values are not going to change I would go with struct. You can add logic (but you can do that with a class also). Go decimal unless you specifically need double.

public struct StdDev
{
    public decimal Val { get; }
    public decimal Dev { get; }
    public decimal Max { get { return Val + Dev; } }
    public decimal Min { get { return Val - Dev; } }
    public bool IsInDev (decimal val)
    {
        return val >= Min && val <= Max;
    }
    public override string ToString()
    {
        return $"{Val} +- {Dev}";
    }
    public StdDev (decimal val, decimal dev)
    {
        Val = val;
        Dev = dev;
    }
}

You could override Equals and GetHashCode.

Answer 2

Here are some other options to what you currently have:

It's up to you to choose and use what is convinient. Don't know if performance is what you are looking for.

Option 1 (if you are using C# 7.0 or later):

How about Tuples ?

You won't have key value pair in tuple but rather items.

Option 2: Like Dan wilson mentioned in the comment to your question make a class with value and error

public class ValueWithDeviation
{
    public double Value {get; set;}

    public double Deviation {get; set;}
}

Then you could have IEnumerable<ValueWithError>

Answer 3

As DanWilson and Cybercop suggest, I'd say that creating a dedicated type ( a class or a struct ) is the way to go, because the value with the deviation can be seen together as a logical object. Also, that will make your life easier when you want to display the values, as you will just need to override the ToString method in your type.

As for the underlying type, it might be more convenient for you to use decimal instead of double , but that depends on your specific project/needs.