Methods and apparatus to categorize media impressions by age

Abstract

Methods, apparatus, and articles of manufacture are disclosed to categorize audience members by age. An example method disclosed herein includes assigning a weight to each audience member record. In the example method, the weight is based on a quantity of audience members in a same age group as the audience member record. The example method includes, at the child nodes, calculating an effective quantity of audience member records based on the weights of the audience member record assigned to the corresponding child node. In the example method, when the effective quantity of audience member records satisfies a minimum leaf size, splitting the corresponding child node into additional ones of the child nodes based on a corresponding child node attribute-value pair. In the example method, when the effective quantity of audience member records does not satisfy the minimum leaf size, designating the corresponding child node as a terminal node.

1 . A method to categorize a media impression, the method comprising:
 generating, by executing an instruction with a processor, a plurality of audience member records by associating ages of audience members from a first memory of a first database proprietor at a first Internet domain to the corresponding activity data obtained from a second memory of a second database proprietor at a second Internet domain;   assigning, by executing an instruction with the processor, a weight to each audience member record, the weight being based on a quantity of audience members in a same age group as the audience member record;   generating, by executing an instruction with the processor, an age-correction model by:
 at an initial node, splitting the plurality of audience member records into child nodes based on an initial attribute-value pair; 
 at the child nodes:
 calculating an effective quantity of audience member records based on the weight of the audience member record assigned to the corresponding child node; 
 when the effective quantity of audience member records satisfies a minimum leaf size, splitting the corresponding child node into additional ones of the child nodes based on a corresponding child node attribute-value pair; and 
 when the effective quantity of audience member records does not satisfy the minimum leaf size, designating the corresponding child node as a terminal node; 
 generating probability density functions to infer probabilities reflecting underrepresented ages in the probability density functions, the probability density functions based on the audience member records assigned to the respective terminal node that includes the underrepresented age samples; and 
 
   correcting, based on the age-correction model, the age characteristic associated with the media impression, the media impression indicative of a person being exposed to media presented by a media presentation device, the correcting to generate more accurate audience measurement information by including ages that are underrepresented in the probability density functions without consuming additional memory and processor resources to monitor additional audience members in the underrepresented ages.   
     
     
         2 . The method as defined in  claim 1 , wherein the assigning of the weights to the plurality of samples corresponding to the different age groups further includes, when the quantity of audience member records grouped within one of the age groups satisfies a first threshold, assigning a neutral weight to the corresponding samples in the one of the age groups. 
     
     
         3 . The method as defined in  claim 1 , wherein the assigning of the weights to the plurality of samples corresponding to the different age groups further includes, when the quantity of the audience member records grouped within one of the age groups satisfies a second threshold, assigning a maximum weight to the corresponding samples in the one of the age groups. 
     
     
         4 . The method as defined in  claim 1 , wherein the assigning of the weights to the plurality of samples corresponding to the different age groups further includes, when the quantity of the audience member records grouped within one of the age groups satisfies a second threshold, assigning a proportional weight between a neutral weight and a maximum weight to the corresponding samples in the one of the age groups. 
     
     
         5 . (canceled) 
     
     
         6 . The method as defined in  claim 1 , further including generating the plurality of audience member records based on survey data obtained from a plurality of panelist and activity data corresponding to the plurality of panelist retrieved from the database proprietor. 
     
     
         7 . An apparatus comprising:
 a record generator to retrieve, from a first memory of an audience measurement entity, a plurality of audience member records corresponding to different age groups, the plurality of audience member records including behavior attributes collected at a second memory of a database proprietor;   a weight calculator to assign a weight to each audience member record, the weight being based on a quantity of audience members in a same age group as the audience member record;   a model builder to:
 at an initial node, split the plurality of audience member records into child nodes based on an initial attribute-value pair; 
 at the child nodes:
 calculate an effective quantity of audience member records based on the weight of the audience member record assigned to the corresponding child node; 
 when the effective quantity of audience member records satisfies a minimum leaf size, split the corresponding child node into additional ones of the child nodes based on a corresponding child node attribute-value pair; 
 when the effective quantity of audience member records does not satisfy the minimum leaf size, designate the corresponding child node as a terminal node; 
 generating probability density functions to infer probabilities reflecting underrepresented ages in the probability density functions, the probability density functions based on the audience member records assigned to the respective terminal node that includes the underrepresented age samples; and 
 
 determine, based on the model, an age characteristic of a person exposed media presented by a media presentation device, the determining to generate more accurate audience measurement information by including ages that are underrepresented in the probability density functions without consuming additional memory and processor resources to monitor additional audience members in the underrepresented ages. 
   
     
     
         8 . The apparatus as defined in  claim 7 , wherein to assign the weights to the plurality of samples corresponding to the different age groups, the weight calculator is to, when the quantity of audience member records grouped within one of the age groups satisfies a first threshold, assign a neutral weight to the corresponding samples in the one of the age groups. 
     
     
         9 . The apparatus as defined in  claim 7 , wherein to assign the weights to the plurality of samples corresponding to the different age groups, the weight calculator is to, when the quantity of the audience member records grouped within one of the age groups satisfies a second threshold, assign a maximum weight to the corresponding samples in the one of the age groups. 
     
     
         10 . The apparatus as defined in  claim 7 , wherein to assign the weights to the plurality of samples corresponding to the different age groups, the weight calculator is to, when the quantity of the audience member records grouped within one of the age groups satisfies a second threshold, assign a proportional weight between a neutral weight and a maximum weight to the corresponding samples in the one of the age groups. 
     
     
         11 . (canceled) 
     
     
         12 . The apparatus as defined in  claim 7 , wherein the record generator is to generate the plurality of audience member records based on survey data obtained from a plurality of panelist and activity data corresponding to the plurality of panelist retrieved from a database proprietor. 
     
     
         13 . A tangible computer readable storage medium comprising machine readable instructions that, when executed, cause a machine to at least:
 retrieve, from a first memory of an audience measurement entity, a plurality of audience member records corresponding to different age groups, the plurality of audience member records including behavior attributes collected at a second memory of a database proprietor;   assign a weight to each audience member record, the weight being based on a quantity of audience members in a same age group as the audience member record;   at an initial node, split the plurality of audience member records into child nodes based on an initial attribute-value pair;   at the child nodes:
 calculate an effective quantity of audience member records based on the weight of the audience member record assigned to the corresponding child node; 
 when the effective quantity of audience member records satisfies a minimum leaf size, split the corresponding child node into additional ones of the child nodes based on a corresponding child node attribute-value pair; and 
 when the effective quantity of audience member records does not satisfy the minimum leaf size, designate the corresponding child node as a terminal node; 
 generate probability density functions to infer probabilities reflecting underrepresented ages in the probability density functions, the probability density functions based on the audience member records assigned to the respective terminal node that includes the underrepresented age samples; and 
   determine, based on the model, an age characteristic of a person exposed to media presented by a media presentation device, the determining to generate more accurate audience measurement information by including ages that are underrepresented in the probability density functions without consuming additional memory and processor resources to monitor additional audience members in the underrepresented ages.   
     
     
         14 . The tangible computer readable storage medium as defined in  claim 13 , wherein to assign the weights to the plurality of samples corresponding to the different age groups, the instructions, when execute cause the machine to, when the quantity of audience member records grouped within one of the age groups satisfies a first threshold, assigning a neutral weight to the corresponding samples in the one of the age groups. 
     
     
         15 . The tangible computer readable storage medium as defined in  claim 13 , wherein to assign the weights to the plurality of samples corresponding to the different age groups, the instructions, when execute cause the machine to, when the quantity of the audience member records grouped within one of the age groups satisfies a second threshold, assigning a maximum weight to the corresponding samples in the one of the age groups. 
     
     
         16 . The tangible computer readable storage medium as defined in  claim 13 , wherein to assign the weights to the plurality of samples corresponding to the different age groups, the instructions, when execute cause the machine to, when the quantity of the audience member records grouped within one of the age groups satisfies a second threshold, assigning a proportional weight between a neutral weight and a maximum weight to the corresponding samples in the one of the age groups. 
     
     
         17 . (canceled) 
     
     
         18 . The tangible computer readable storage medium as defined in  claim 13 , further including instructions that, when execute cause the machine to generate the plurality of audience member records based on survey data obtained from a plurality of panelist and activity data corresponding to the plurality of panelist retrieved from the database proprietor.