Page 161 - Leaders in Legal Business - PDF - Final 2018
P. 161
also illuminate the data. Furthermore, software can depict those features of the data in graphics,
such as histograms, density plots, scatterplots, and bar charts.
Correlations between variables in the data.
It may be useful for legal managers to see how one variable (an element of data tracked
for every associate, client, lawyer, or whatever) moves up or down in relation to the average
when another variable changes. Thus, for instance, the software can show the correlation
between the number of matters worked on by a lawyer and billable hours reported. A correlation
tells you whether there is an association between two variables, how strong it is, and in what
direction the variables move. It is a positive correlation if both numbers move in the same
direction (such as higher billable hours and higher bonuses); it is a negative correlation if the
numbers move in the opposite direction (such as higher billable hours and lower psychological
well-being).
Comparisons of averages and differences.
Several statistical tools can detect whether the difference between two or more numbers
has significance mathematically. So, for example, there are many techniques to tell whether the
average billable hours in a year between two offices of a law firm vary enough for managers to
consider intervening and taking some action. These tools, such as ANOVA and the Student’s T-
Test, help to determine whether variations are important enough to deserve discussion.
Measures of inequality.
Managers of lawyers may want to assess the quality of a set of numbers, such as bonus
distributions. Along with the well-known Gini coefficient, several other measures allow software
to put a number on inequality and even pinpoint where in the set of numbers the actual data
diverges from theoretical equality. These analytics help managers explain their decisions and
make better decisions in the first place, if equality is sought.
Understand influence of variables and make predictions.
A whole family of regression tools goes beyond correlations. If, for example, a firm
wants to predict the estimated amount of fees to be paid to it during the coming year, it could run
a linear regression. The software would then point out which of the variables was more
influential in predicting total fees paid and how much of the total fee paid is accounted for by the
variables. One of the best-known techniques is multiple linear regression. It makes some
assumptions about the relationships between whatever value is being predicted and the variables
that are associated with it (e.g., level of the person retaining the firm, presence of a law
department, range of practice groups involved, and years as a client).
The regression algorithms generate a “model.” Once you have a model, you can extract
information from it. A model often takes in data and makes predictions regarding new cases,
clients, or matters. Think of a model as the software learning on a “training set” of data that has
been labeled, such as settled for less than $10,000 or not) and applying that learning to predict
something (maybe total fees) for a new case or example. With multiple regression, naïve Bayes
147
such as histograms, density plots, scatterplots, and bar charts.
Correlations between variables in the data.
It may be useful for legal managers to see how one variable (an element of data tracked
for every associate, client, lawyer, or whatever) moves up or down in relation to the average
when another variable changes. Thus, for instance, the software can show the correlation
between the number of matters worked on by a lawyer and billable hours reported. A correlation
tells you whether there is an association between two variables, how strong it is, and in what
direction the variables move. It is a positive correlation if both numbers move in the same
direction (such as higher billable hours and higher bonuses); it is a negative correlation if the
numbers move in the opposite direction (such as higher billable hours and lower psychological
well-being).
Comparisons of averages and differences.
Several statistical tools can detect whether the difference between two or more numbers
has significance mathematically. So, for example, there are many techniques to tell whether the
average billable hours in a year between two offices of a law firm vary enough for managers to
consider intervening and taking some action. These tools, such as ANOVA and the Student’s T-
Test, help to determine whether variations are important enough to deserve discussion.
Measures of inequality.
Managers of lawyers may want to assess the quality of a set of numbers, such as bonus
distributions. Along with the well-known Gini coefficient, several other measures allow software
to put a number on inequality and even pinpoint where in the set of numbers the actual data
diverges from theoretical equality. These analytics help managers explain their decisions and
make better decisions in the first place, if equality is sought.
Understand influence of variables and make predictions.
A whole family of regression tools goes beyond correlations. If, for example, a firm
wants to predict the estimated amount of fees to be paid to it during the coming year, it could run
a linear regression. The software would then point out which of the variables was more
influential in predicting total fees paid and how much of the total fee paid is accounted for by the
variables. One of the best-known techniques is multiple linear regression. It makes some
assumptions about the relationships between whatever value is being predicted and the variables
that are associated with it (e.g., level of the person retaining the firm, presence of a law
department, range of practice groups involved, and years as a client).
The regression algorithms generate a “model.” Once you have a model, you can extract
information from it. A model often takes in data and makes predictions regarding new cases,
clients, or matters. Think of a model as the software learning on a “training set” of data that has
been labeled, such as settled for less than $10,000 or not) and applying that learning to predict
something (maybe total fees) for a new case or example. With multiple regression, naïve Bayes
147
