A number of steps were undertaken in the development and validation processes of the two age-appropriate children’s cooking competence measures. Items were selected for inclusion for the content of the measures from a review of children’s cooking recommendations that were mapped to their age-appropriate developmental skills , and were reviewed by an expert panel. The measure was designed in line with a published perceived motor competence measure  and characters performing the cooking skills were illustrated by a graphic designer. Next initial face validity was established by the research team and a primary school teacher and the measures were piloted. The psychometric properties of the measures were assessed using Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) and face validity of the final structures of the two measures were established. Following this, construct validity, convergent and discriminant, and internal consistency reliability were assessed. Temporal stability reliability of both measures was then examined. Finally, for the older measure, responsiveness to change was established. The details of this process can be found in the following sections.
2.1 Item selection
The cooking skills that children should be learning at different ages were obtained through a critical evaluation of publically available children’s recommendations and the addition of new recommendations based on children’s developmental skills . From this review, for the two Cooking Competence measures, 14 cooking skills for 8 - 12 year olds (CooC11) and 10 skills for 6-7 year olds (CooC7) were identified as being frequently occurring and culturally neutral, i.e. a cooking skill was not specific to one culture. The items were selected to ensure they were both developmentally appropriate and relevant . An expert panel including an educational researcher, a primary school teacher, an early year’s educator, a movement scientist and two Home Economists, with a minimum of 10 years’ experience in their respective fields, reviewed the selected skills for age appropriateness and level of difficulty. An age range was proposed for each skill and skills were then ranked in order of difficulty from easiest to hardest, see Table 1.
Table 1: Cooking skills identified for younger and older children
a – measuring liquids and weighing were separated in the measure; b - tin opener was placed in the younger age before the expert panel review moved it into the 9+ age category; c – blender replaced mixer as mixer was more associated with baking as opposed to cooking; d – oven and microwave were separated in the measure; e – Using the stove/hob was added as a means of factoring in the use of a cooker/cooker top for those that may not have an oven/as a means of trying to include stirring over heat
2.2 Development and implementation of measure
The design of the children’s perceived cooking competence measure was based on a published perceived motor competence measure . However, in the cooking skills measure the child was asked first whether they engaged in a particular cooking skill (in line with Lavelle et al. ), before they rated their level of competence. This aims to reduce positive illusory and social desirability biases . In the measure, each skill was illustrated as a child-friendly character performing the skill. Using an iterative process, the child characters were drawn by a graphic designer in consultation with a chef and reviewed by the research team for accuracy and suitability. The illustrations provided a visual ‘cue’ to the cooking skill as some cooking terminology relating to skills may not be familiar to the children. In line with Barnett et al. , the child is shown an image of a child, boys are presented with images of boys performing the skills and girls are shown images of girls. This promotes a peer modelling effect, as it is argued that a child is more likely to relate to a character that is more like themselves [41, 42]. The child is asked whether they do the skill shown (see figure 1 as an example). If the child responds yes, then they are shown two more images of the child performing the skill, one performing it well and the other performing it poorly. The child is then asked which image represents their perceived level of competence on a five point Likert scale.
The five response options result in five possible levels of competence for each skill (see figure 2 as an example). However if the child responds that they do not perform that particular skill, they move on to the next skill.
The cooking skills are presented in ascending level of difficulty as rated by the expert panel. Additionally, the sequence of presentation of ‘good’ competence of a skill alternated in position on the page with ‘poor’ competence of a skill [29, 30]. Each child completes the measure individually. However, if a child’s literacy levels were not at a sufficient level that they could read the questions, then the researchers assisted the child by reading out the questions so that the child could complete the task.
2.3 Piloting and Initial Face Validity
The measures were reviewed by the research team and a primary school teacher for language, readability and literacy levels . Based on the feedback minor amendments were made to the language, such as changing ‘in between A and B’ to ‘A bit like A and B,’ as it was suggested that children would interpret the original phrasing to mean physically in between the two characters. Additionally, the font size of the text was increased for the younger age group. Furthermore, the characters’ expressions were all changed to neutral, so that the children would not choose their responses based on how happy or sad they were feeling but on their perceived level of competence. Thus, the research team assessed the measures for initial face validity.
The designed measures were also piloted with a number of children of differing ages . This piloting allowed the research team to assess the usability, length of time of completion, enjoyability as well as further face validity such as recognition of the skills and differentiation between the ‘good,’ and ‘poor’ performance of the skill.
Further validation assessments were undertaken for both the older age measure (CooC11) and the younger age measure (CooC7), which will be detailed in the following section. For both measures, endpoint user feedback from both the children and teachers were received. Five teachers and three teaching assistants, from the recruited samples of children, provided their perceptions around the measures including the suitability, usability and length. Each class in these samples were asked about their experiences using the measure, whether they liked this type of activity, about the characters and if there was anything they would change. Additionally, informal qualitative feedback was gathered across all samples. Prior to data analysis, where necessary, items were reversed coded so that a higher score indicated greater perceived cooking competence for all items. All analyses were conducted using IBM SPSS Statistics v25 and IBM SPSS Amos v25, with a significance level of 0.05.
2.4 CooC11 (8-12 year olds)
2.4.1 Participants and Procedure
Sample 1: Data from 469 primary school children aged 10 – 11 years completed baseline measurements as part of a larger study (Project Daire) . For this sample, 50.32% were female. Schools from both rural and urban areas with varying socioeconomic levels were included. Data was collected in February-March 2019.
Sample 2: Children (N=38) between the ages of 8-9 years and 10-11 years (two year groups in the primary school system in Northern Ireland) were recruited. Children from one primary school were recruited for this study. For this sample, 52.6% were female. Children in this sample completed the measure at two time points two weeks apart in May 2019.
Sample 3: Children (N=32) between the ages of 10-12 years who participated in a one week cooking camp intervention in August 2019. In this sample, 78.1% were female. These children completed the measure before and after the cooking camp intervention.
2.4.2 Psychometric testing, Validation and Data Analysis
Exploratory Factor Analysis
Sample 1 was randomly split to conduct EFA and CFA, with 269 children included in the EFA. EFA (maximum likelihood) with direct oblimin rotation was used. This oblique rotation was used as it was believed that factors would be related . Sample adequacy was assessed using Kaiser-Meyer-Olkin (KMO) value  and Bartlett’s Test of Sphercity . Factors were assessed using Eigenvalues greater than 1  and a minimum of 3 items per factor . Items were removed based on communalities and factor loadings.
Confirmatory Factor Analysis and Face Validity
The remaining 200 randomly selected children from sample 1 were used for CFA. The final model identified by the EFA was assessed as a confirmatory factor analysis with maximum-likelihood estimation, using IBM SPSS Amos v25. The following fit statistics were used to assess the model :
- Chi-square (χ2) – A non-significant chi-square value (p > 0.05) which is two or three times larger than its value divided by the degrees of freedom (df) at its maximum indicates that the model can be accepted.
- Root Mean Square Error of Approximation (RMSEA) – A preferred value is 0.05 or less.
- Comparative Fit Index (CFI), Normed-Fit Index (NFI), Tucker-Lewis Index (TLI) – For these indices a value of 0.90 or greater indicates that the model can be accepted.
To establish face validity of the measure structure, five researchers in the areas of food, nutrition, health psychology, Home Economics and human movement science, reviewed the final model and factor structure. Cooking skills in each factor were assessed upon general relation in cooking as well as underlying developmental skills including fine and gross motor skills in addition to numeracy, literacy and safety considerations. All items were assessed to ensure they measured what they claimed to measure.
Construct Validity – Convergent and Discriminant Validity
Sample 1 was used for Construct validity. Convergent validity shows that measures are valid by identifying a relationship with an existing similar measure using correlation analysis. As there are no similar children’s measures to establish convergent validity, the cooking method section of an adult measure was used . This measure has not been used previously with children due to the levels of literacy required. However, as the current sample is at the older end of the age range for the measure, the research team decided to include the measure as a means of establishing some level of convergent validity. Additionally, due to the lack of measurements available, the children were asked whether they help their parents making the dinner. It was expected that those who assist with dinner preparation would have a reported higher cooking competence. Due to a large number of children answering ‘sometimes’ (N=298) and a small number answering ‘always’ (N=35), compared with those answering ‘never’, ‘sometimes’ and ‘always’ were combined to those who ‘help with dinner.’ And due to the large difference between the two response categories, a random selection of those who responded ‘help with dinner’ were selected to compare against never. This ensured that there was a relatively equal number of participants in each group for the one-way ANOVA.
Internal Consistency Reliability
Internal consistency reliability was used to examine agreement between the items in a scale. Cronbach’s Alpha was used to assess internal consistency reliability. A value of 0.7 or higher shows good reliability . Sample 1 and 2 were used to establish internal consistency of the measure.
Sample 2 was used to assess Temporal Stability of the measure. The temporal stability of the scales was examined using the Intraclass Correlation Coefficient (ICC). This illustrates the level of agreement between item answers over time. A stronger ICC indicates greater agreement, suggesting greater temporal stability. Moderate reliability is seen with an ICC value of 0.50–0.75, good reliability is a value of 0.75–0.90, while a value of greater than 0.90 suggests excellent reliability .
Responsiveness to Change
Sample 3 was used to assess the responsiveness to change of the measure, a further indication of validation . This was established through investigating changes in the measure scores before and after the children receive a cooking focused intervention using T-tests.
2.5 CooC7 (6-7 year olds)
2.5.1 Participants and Procedure
Sample 4: Data from 514 primary school children aged 6 - 7 completed baseline measurements as part of a larger study (Project Daire) , are used as Sample 4. For this sample, 48.63% were female. Schools with varying socioeconomic levels and from both rural and urban areas were included. Data was collected in February-March 2019.
Sample 5: Children (N=13) between the ages of 6 – 7 years old were recruited as part of Sample 5 from the same school as sample 2. In this sample, 46.2% were female. Children in this sample completed the measure at two time points two weeks apart in May 2019.
2.5.2 Psychometric testing, Validation and Data Analysis
The same criteria as in 2.4.2 were used for testing the CooC7 measure. The samples used, and differences in analysis to CooC11 are detailed below.
Exploratory Factor Analysis
Sample 4 was randomly split to conduct Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA), with 314 children included in the EFA.
Confirmatory Factor Analysis and Face Validity
The remaining 200 randomly selected children from Sample 4 were used for CFA. The same procedure was used for the CFA and face validity as in section 2.4.2.
Construct Validity – Discriminant Validity
Sample 4 was used for construct validity. There are no similar children’s measures to establish convergent validity and the cooking method section of the adult measure  is above the literacy and cognitive capacity of this age group. Due to the lack of measurements available, the children were asked whether they help their parents making the dinner, with the expectation again that a higher cooking competence score would be seen in those that help with dinner preparation. The responses for the children were ‘never’, ‘sometimes’ or ‘always’. Comparisons between children in the 3 categories were conducted using an ANOVA with Scheffe post hoc analysis due to differences in numbers in the groups.
Internal Consistency Reliability and Temporal Stability
Sample 4 and 5 were used to establish internal consistency of the measure. Sample 5 was used to assess Temporal Stability of the CooC7.
2.6 Ethical Considerations
All schools partaking in the research (Samples 1, 2, 4 and 5) signed and returned a memorandum of understanding. An opt-out parental consent system was implemented. In sample 3, due to the nature of the intervention and the demand for places, an opt-in system was used. In all samples, parents were made aware that they were not obliged to allow their child to take part in the study and that they could withdraw their child at any time point up to data analysis without reason or consequence. Additionally, the children were made aware that they did not have to take part. The research was conducted in accordance with the Declaration of Helsinki. Ethical approval was received from The School of Social Sciences, Education and Social Work Ethics Committee, Queen’s University Belfast (Reference number 038_1819) for Samples 1 and 4 and from The School of Biological Sciences Ethics Committee, Queen’s University Belfast (0519/LavelleFA, 0519/LavelleFB), for Samples 2, 3 and 5.