To gain experience in the process of designing an attitude measure, we created a scale to measure self esteem (appendix 1). The history of this can be read in the report ‘The Development of a new Scale to Measure Attitudes to Self’. Interest in the topic led us to question ‘what factors cause self esteem to differ so widely in different individuals’. Research highlighted the importance of parenting style experienced in childhood.
It was decided to focus upon Baumrind’s three classifications of Authoritative, Authoritarian and Permissive parenting styles. We decided to create another scale, this time measuring parenting styles, this would be administered with the first test and the results correlated. We would therefore be able to see if there was any relationship between the two issues. Before it was possible to create our own questionnaire, it was necessary to consult an existing form measuring the same thing. We examined Buri et als Parental Authority Questionnaire (PAQ), used in their 1991 study (appendix 2). As such we were able to see how the questions were formed to address each of the three different styles.
We divided into three groups and in a brain storming session generated statements referring to each of the parenting styles. The resultant 28 authoritative, 18 authoritarian and 27 permissive statements were collated from the three groups (appendix 3). It was necessary to ensure that the statements were clear and simple, contained: no complex language; less than twenty words; only one Idea/thought; no double negatives and no definites, for example, ‘always’ and ‘never’. Buri et als PAQ was answered. The range was calculated for the results (appendix 4). Questions with the widest range were selected for our PAQ.
The resultant 80 statements were randomised (appendix 5) so that the statements for the different parenting styles were separated, also this would prevent the participants from realising the concepts being tested. A five point scale was added to the questions, where 1= strongly disagree, and 5 = strongly agree. This was piloted on 10 participants (us): nine females and one male. The age ranged from 18 – 27 years.
The range and standard deviation were calculated for the results (appendix 6). Any questions with a range of two and below were discarded. To reduce the ‘item pool’ even further, statements which did not have: face validity, that is, did not ‘look’ like they measured the relevant idea or content validity, that is, did not measure the relevant idea; were discarded. Statements containing: ambiguity and more than one idea were also removed from the measure as they can create confusion for the participant.
Statements which could have elicited socially desirable responses as opposed to true responses, that is, participants would not agree with the statement as it could reflect him/her in a negative light were removed, otherwise this would damage the accuracy of the results. Offensive statements were removed because it would not be ethical to insult the participant and also it is necessary for the participant to complete the questionnaire for us to use the results. It was important that the term ‘caregiver’ as opposed to ‘parent’ was used to ensure that the PAQ applied to everyone.
It was ensured that thirty questions remained, 10 of the statements referred to each of the parenting styles. The statements which had been placed into order of the best range were returned to their original random order and were renumbered 1-30 (appendix 7). In our groups, we then had to write an opening paragraph to introduce and explain the measure; a conclusion to thank the participants and a scale to measure the attitudes.
To do this it was necessary to research existing scales. It was decided that a Likert, five point scale would be the most acceptable because it would encourage the participant to use the whole scale, whereas with a seven (or more) point scale, the participants may only use the middle values. To indicate responses it was decided that the participant should circle the number relating to their choice, this would be simple and would not irritate the participant. The introduction, scale and conclusion were typed up with the questions creating the finished questionnaire (appendix 7). Next the PAQ was administered with our self-esteem measure. The data was collated (appendix10). To score the data for our PAQ, the scores given for the statements for each parenting style were totalled, resulting in three overall scores for each of the styles (appendix 11).
The largest of which was taken as the parenting style specific to the participant. For our self esteem measure, the scores for all of the statements referring to low self esteem were negated, that is a score of 5 became a 1 and vice versa. Statements referring to high self esteem were left unchanged, the scores for the entire questionnaire were totalled resulting in one self esteem score. Buri et als PAQ (appendix 2) and Coopersmith’s ‘Self-Esteem Questionnaire I’ (appendix 2) were scored according to their instructions (appendix 10).
Participants 2, 3 and 18 revealed tied highest scores, that is, the highest score on the original test was taken to be the relevant parenting style, however in these cases, as the highest score was equal for two parenting types it was not possible to select the relevant style, as such they had to be discarded. In the previous study, the creation of a scale to measure self esteem a split half reliability test found a correlation of 0.45. A split half reliability measure was carried out on the original test for our PAQ (appendix 14).
This was done by taking the first five statements referring to each of the parenting styles and correlating it using Spearmans Rho with the remaining five questions for each of the parenting styles (appendix 13). A correlation of 0.648 was found for the authoritative statements. A correlation of 0.537 was found for the authoritarian questions; whilst a correlation of 0.405 was found for the permissive statements. The British Psychological Society (BPS) recommend a correlation of 0.7 for a reliable test.
A test-retest reliability measure was carried out on the scored data for our PAQ (appendix 15). A correlation of 0.877 was found for the authoritative statements. The authoritarian statements had a test-retest correlation of 0.841. Whilst a correlation of 0.814 was found for the permissive statements. Coolican argues that for test-retest reliability a correlation of 0.9 should be found, this would suggest that our test does not have good enough test-retest reliability. However according to the BPS this is satisfactory.
Buri et als PAQ was used as an alternate form allowing us to perform a reliability measure. A correlation of 0.117 was found for the authoritative statements. A correlation of 0.463 was found for the authoritarian questions. The permissive questions found a correlation of 0.096 (appendix 16). Furthermore Coopersmiths ‘Self-Esteem Questionnaire I’ was also used as an alternate form with our self esteem questionnaire. A reliability test found a correlation of 0.075. (appendix 16)
Three scatter diagrams correlating self esteem with the different parenting styles were created, allowing us to see if a correlation existed. There were only 2 participants who appeared to have experienced permissive parenting styles in childhood, this is not enough data and so this parenting style was discarded. It was necessary to calculate a Spearman’s Rho to determine if these results were significant (appendix 17).
Self esteem was correlated against the two remaining parenting styles (appendix 18). However this was not found to be significant. In the last study a problem with the measure was highlighted. The self esteem measure appeared to test more then one factor and as such a low split half correlation of 0.45 was found, this it was decided, was skewing our results. Factor analysis was carried out on the data by creating a matrix that correlated each item against itself and every other item. For each question we counted the number of times a correlation occurred that was greater than 0.5 and -0.5.
At the bottom of the column we wrote the number of correlations plus one for the question itself. Next we had a look at the specific questions that correlated, generally these appeared to address an obvious factor. Some of the questions however did not appear to be asking about a similar factor and so were discarded. We identified seven factors, these were assigned appropriate names, these included sociability, confidence, irritability/anxiety, sensitivity, expressiveness (in terms of emotion), arrogance/modesty, and shy/nervous. Some questions however did not correlate with anything and were also discarded (appendix 19)..
We went on to correlate (using Spearman’s Rho) each of the seven factors against the two remaining parenting styles (appendix 20). The results were checked for significance in the correlation tables. Using each of these eight pairs of correlations (self esteem and the seven factors). We were able to perform a Mann Whitney U on the two sets of self esteem scores taken from the remaining authoritative and authoritarian parenting styles. Allowing us to see if there is a correlation between parenting styles and self esteem (or the seven factors)(appendix 20).
RESULTS
The results for our PAQ (original test) and our Self-Esteem measure (original test) were collated, this data was scored (appendix 10). On the self esteem measure, questions referring to high self esteem were unchanged, however for those referring to low self esteem the scores were negated, so that a 5 became a 1 and vice versa. On our PAQ, the scores for the questions referring to each of the individual parenting styles were totalled, resulting in three separate scores, an authoritative, an authoritarian and a permissive score. For the PAQ the highest number was taken. Participants 2, 3 and 18 had equal highest score for two different parenting styles, they were therefore discarded. A summary of these four scores can be seen in appendix. Tables 1-3 below show a summary of the specific parenting style experienced by the participant and his/her self-esteem score.